This PR adds a `--format=bash` option to `spack commands` to
auto-generate the Bash programmable tab completion script. It can be
extended to work for other shells.

Progress:

- [x] Fix bug in superclass initialization in `ArgparseWriter`
- [x] Refactor `ArgparseWriter` (see below)
- [x] Ensure that output of old `--format` options remains the same
- [x] Add `ArgparseCompletionWriter` and `BashCompletionWriter`
- [x] Add `--aliases` option to add command aliases
- [x] Standardize positional argument names
- [x] Tests for `spack commands --format=bash` coverage
- [x] Tests to make sure `spack-completion.bash` stays up-to-date
- [x] Tests for `spack-completion.bash` coverage
- [x] Speed up `spack-completion.bash` by caching subroutine calls

This PR also necessitates a significant refactoring of
`ArgparseWriter`. Previously, `ArgparseWriter` was mostly a single
`_write` method which handled everything from extracting the information
we care about from the parser to formatting the output. Now, `_write`
only handles recursion, while the information extraction is split into a
separate `parse` method, and the formatting is handled by `format`. This
allows subclasses to completely redefine how the format will appear
without overriding all of `_write`.

Co-Authored-by: Todd Gamblin <tgamblin@llnl.gov>

* Revert "Travis: use --concurrency=multiprocessing only on build tests (#6872)"

This reverts commit 596d4637.

* Removing 'coverage combine' in test script

According to what was discovered in #6887, one of the problems is
calling 'coverage combine' twice without the '-a' flag. This removes
the first call within our test scripts.

On a local workstation, it seems that tracking multiple processes during
coverage may result in malformed coverage reports for unit tests and not
for build tests.

Given that multiple processes make a difference in coverage mainly for
build tests, try to disable the tracking for unit tests to see if we get
more stable coverage results.

* Porting: substitute nose with ytest

This huge commit substitutes nose with pytest as a testing system. Things done here:

* deleted external/nose as it is no longer used
* moved mock resources in their own directory 'test/mock/'
* ported two tests (cmd/find, build_system) to pytest native syntax as an example
* build_environment, log: used monkeypatch instead of try/catch
* moved global mocking of fetch_cache to an auto-used fixture
* moved global mocking from test/__init__.py to conftest.py
* made `spack test` a wrapper around pytest
* run-unit-tests: avoid running python 2.6 tests under coverage to speed them up
* use `pytest --cov` instead of coverage run to cut down testing time

* mock/packages_test: moved mock yaml configuration to files instead of leaving it in the code as string literals

* concretize.py: ported tests to native pytest, reverted multiprocessing in pytest.ini as it was creating the wrong report for coveralls

* conftest.py, fixtures: added docstrings

* concretize_preferences.py: uses fixtures instead of subclassing MockPackagesTest

* directory_layout.py: uses fixtures instead of subclassing MockPackagesTest

* install.py: uses fixtures instead of subclassing MockPackagesTest

* optional_deps.py: uses fixtures instead of subclassing MockPackagesTest

optional_deps.py: uses fixtures instead of subclassing MockPackagesTest

* packages.py: uses fixtures instead of subclassing MockPackagesTest

* provider_index.py: uses fixtures instead of subclassing MockPackagesTest

* spec_yaml.py: uses fixtures instead of subclassing MockPackagesTest

* multimethod.py: uses fixtures instead of subclassing MockPackagesTest

* install.py: now uses mock_archive_url

* git_fetch.py: uses fixtures instead of subclassing MockPackagesTest

* hg_fetch.py: uses fixtures instead of subclassing MockPackagesTest

* svn_fetch.py, mirror.py: uses fixtures instead of subclassing MockPackagesTest
repo.py: deleted

* test_compiler_cmd.py: uses fixtures instead of subclassing MockPackagesTest

* cmd/module.py, cmd/uninstall.py: uses fixtures instead of subclassing MockDatabase

* database.py: uses fixtures instead of subclassing MockDatabase, removed mock/database

* pytest: uncluttering fixture implementations

* database: changing the scope to 'module'

* config.py: uses fixtures instead of subclassing MockPackagesTest

* spec_dag.py, spec_semantics.py: uses fixtures instead of subclassing MockPackagesTest

* stage.py: uses fixtures instead of subclassing MockPackagesTest. Removed mock directory

* pytest: added docstrings to all the fixtures

* pytest: final cleanup

* build_system_guess.py: fixed naming and docstrings as suggested by @scheibelp

* spec_syntax.py: added expected failure on parsing multiple specs closes #1976

* Add pytest and pytest-cov to Spack externals.

* Make `spack flake8` ignore externals.

* run-unit-tests runs spack test and not pytest.

* Remove all the special stuff for `spack test`

- Remove `conftest.py` magic and all the special case stuff in `bin/spack`

- Spack commands can optionally take unknown arguments, if they want to
  handle them.

- `spack test` is now a command like the others.

- `spack test` now just delegates its arguments to `pytest`, but it does
  it by receiving unknown arguments and NOT taking an explicit
  help argument.

* Fix error in fixtures.

* Improve `spack test` command a bit.

- Now supports an approximation of the old simple interface
- Also supports full pytest options if you want them.

* Use external coverage instead of pytest-cov

* Make coverage use parallel-mode.

* change __init__.py docs to include pytest