Since packages store etags on disk directly, the local install feature
can no longer depend on this; we instead port the etag feature back to
local providers only.

This patch makes it so that whenever the rebar3 shell has a new
configuration for an application that is already running and would be
restarted (without risking the stability of the node or functionality of
rebar_agent), we stop and restart the app.

When adding the 'TEST' macro to the test profile, we mistakenly sourced
the erl_opts values from the base profile rather than the test profile
itself.

This means that in cases where the base profile set an option such as
'no_debug_info' and a profile overrode it with 'debug_info', the default
options would get injected within the test profile, and broke the
precedence rules, yielding incompatible values.

This patch fixes things by adding the macro to the values sourced from
the test profile itself, fixing the issue.

Let's bundle this with another PR

This will at least display the script that failed with the stacktrace
before failing as usual, but without altering the return value.

This should make the common failure path more user-friendly without
breaking any existing behaviour that may have relied on the script file
working. If there's any unexpected side-effect, it will be visual only
rather than blocking full builds if ?ABORT were used.

Some still remain from erl_type calls from Dialyzer, but most of them
are handled.

Decided to just ignore rebar_alias since playing with the type specs of
abstract code format is just a nightmare and hard to do cross-versions I
guess.

We can probably still move ahead without a cache if we must

We've had multiple tickets opened because of unclear PROXY settings when
the scheme is missing form the URI. To be helpful, we instead add them
dynamically whenever they're missing.

Example issues:
- https://github.com/erlang/rebar3/issues/1747
- https://github.com/erlang/rebar3/issues/1697

It makes things worse than if it weren't there.
Fixes issue #1726

This reverts commit 0f7e6c31e9.

This reverts commit 4ad1db9733.

This reverts commit 3f8dd5eace.

The way arguments are passed by the alias provider is not compatible
with all rebar providers/commands. Especially the release (relx)
provider does not like getting its arguments as a plain string. It
expects its arguments in a pre-parsed format as returned by
getopt:parse/2.

Other commands, e.g. eunit, seem to be fine with both ways of argument
passing. Therefore, this fix changes the alias provider argument passing
to the getopt format in general.

find_apps didn't read config files so no hooks were in the app_infos,
and now that hooks are being done rebar needs plugins to be able
to run clean plugin hooks in deps.

Although fairly irregular, the presence of a subdirectory in ebin/ while
running rebar3 escriptize would crash the run and generate a dump.

Since ebin subdirectories are generally not supported (adding a path for
an application only adds ebin/ and no subdirectories to the VM's path
set), we just silently ignore the directory altogether.

Fixes #1693

Rather than the caller having to think of what to purge or not, use
erlang:check_process_code/2 to detect if the caller (rebar3) may die
because of the operation.

If so, do a soft purge with a conditional delete instead of a hard purge
with a mandatory delete.

Prevents the killing of a plugin with itself

- Uses the code at https://github.com/tsloughter/rebar_alias and brings
  it within rebar3
- adds safety checks to prevent redefining built-in commands or obvious
  circular dependencies between commands (indirect circular deps are
  still possible)
- adds tests
- adds a systest to ensure no clash with the existing plugin

Has no ill effect as tested on Win10

The path reloading of plugins had been fixed properly, but the problem
is that the paths it was using to re-load only considered the current
compile step, rather than the overall state of plugins. As such, the
reloaded paths after plugin compilation only reloaded the *latest*
plugin and not the other ones.

This fix forces the addition of all built plugin paths to the code paths
after a plugin compile job is run. This ensures that the path is clean
for initial plugin deps (only add those that are required), and is
re-made total after the fact (add all the plugins possible).

This commit also includes a system tests suite that can be run
optionally; the problem with this plugin mechanism was impossible to
find through mocked dependencies, and a working counterexample was
provided to us. The systest suite can be run against real projects
without conflict to make sure no regressions are hit.

Adds an option (-m, --min_coverage, or {cover_opts, {min_coverage,X}})
to the 'cover' command, where the value is an integer between 0 and 100.

If the total coverage during the analysis is below the value received,
the command will fail with output like:

    ===> Requiring 64% coverage to pass. Only 62% obtained

If the rate is correct, the command silently passes as it currently
does.

This feature allows to enforce code coverage standards in a project if
desired.

This commit adds a common test hook along with the cth_readable stuff
whose role is to track failing test cases, and create a test
specification out of them.

The test specification is dumped on disk at
_build/<profile>/logs/retry.spec and can be accessed by calling
'rebar3 ct --retry'. This will auto-load the spec file if it can be
found and re-run the failing cases.

If any other argument is found on the list specifying tests, the
'--retry' argument is ignored.

All code for this is marked as experimental in case we end up (keeping
and then) dropping the feature.

This commit adds a common test hook along with the cth_readable stuff
whose role is to track failing test cases, and create a test
specification out of them.

The test specification is dumped on disk at
_build/<profile>/logs/retry.spec and can be accessed by calling
'rebar3 ct --retry'. This will auto-load the spec file if it can be
found and re-run the failing cases.

If any other argument is found on the list specifying tests, the
'--retry' argument is ignored.

All code for this is marked as experimental in case we end up (keeping
and then) dropping the feature.

The Erlang compiler runs based on a global state built from currently
loaded libraries and the configured code path that is available. For
this reason, the rebar3 compiler job unloads all plugin paths before
calling the Erlang compiler.

However, this causes a problem when an application uses a custom
resource handler with a dynamic version in their .app.src file since the
plugin that can be used to find the version has been unloaded.

Fortunately, the compile phase that runs the version handling is
distinct from the phase that uses the Erlang compiler. This patch fixes
the problem by re-loading the plugins' paths in memory before generating
the .app file, and before unloading them afterwards.

It appears that unloading them is unnecessary because the hooks after
that will re-load them, but it is likely better to play it safe with
that global state and clean up after ourselves. It offers better
protection for future changes.

Fixes #1657