alongside their source (ie, if `some_tests` is a directory that
contains test suites beams resulting from compiling them will be
placed in `some_tests` in the appropriate `_build` directory

generate line by line coverage reports

- Allow 'rebar3 help' to work and have it point to 'rebar3 help
  <template>'
- show a 'template not found' message when a template is indeed not
  found rather than crashing.

to the appropriate `_build` dir

matching directory paths

run subsets of test suites

Much clearer semantics now. All lists are treated as proplists, meaning
we want to:
1) allow duplicates (providers have to avoid them if they must)
2) preserve order of elements that compare equal (`a == {a, val}`)
   through a stable sort (so if `{a, b}` comes before `a`, we keep
   `{a, b}` first in the list
3) In two lists of attributes requiring a merge, we always give the
   'new' profile a priority to override the default one.

With the new priority order, and knowing Relx processes things in
reverse already (possibly building a dict internally), we should flip
our options around to keep them correct.

Rather than using the stdlib lists:umerge, we expand it to allow fuzzy
matching on tuples vs. vals (`key` vs. `{key,val}`) with an overriden
sort order so that two tuples or values comparing equal get a priority
on the newest profile.

This is a partial fix for #287 -- this current patch should be followed
by a relx update to take options in order (as if they were a proplist)
to complete it.

Also expose it in rebar_api

The option {deps_error_on_conflict, true} will make it so conflicts in
deps being fetched interrupts the operation rather than just display a
warning.

Defaults to `false'.

The deps are sorted and merged, but the merge function merges lists, not
elements. This yields deps that are duplicated and ran for multiple
times.

We first add proper sorts so the keymerge is guaranteed to be fine, and
then do a dedup run to get rid of duplicates if they happen to be.

Breaking up initial call to parse from the ones deep inside the provider
parsing to do smarter namespace detection, added 'as' the ability to
look into these also, and cleaned up the code a whole lot that would
depend on implicit assumptions.

A side-effect is that 'do' is now valid for all namespaces, although it
can be overriden.

error before:

===> Uncaught error: {'EXIT',
                             {badarg,
                              [{io,format,
                                [<0.23.0>,
                                 "\e[0;36m===> Skipping template ~p, due to presence of a custom template at ~s~n~n\e[0m",
                                 ["/home/mariano/.config/rebar3/templates/fn_app/fn_app.template"]],
                                []},
                               {rebar_templater,prioritize_templates,2,
                                [{file,"src/rebar_templater.erl"},{line,330}]},
                               {rebar_templater,find_templates,1,
                                [{file,"src/rebar_templater.erl"},{line,274}]},
                               {rebar_templater,new,4,
                                [{file,"src/rebar_templater.erl"},{line,47}]},
                               {rebar_prv_new,do,1,
                                [{file,"src/rebar_prv_new.erl"},{line,44}]},
                               {rebar_core,do,2,
                                [{file,"src/rebar_core.erl"},{line,129}]},
                               {rebar3,main,1,
                                [{file,"src/rebar3.erl"},{line,46}]},
                               {escript,run,2,
                                [{file,"escript.erl"},{line,752}]}]}}

root project

get correct compile paths included in module info

This adds some more convolution to command parsing, unfortunately, but
makes some cases clearer (`rebar3 as as as as` is no longer valid). It
unfortunately disallows using $REBAR_PROFILE along with `as` in a rebar
command.