* no clean -t option, use 'spack test remove'

* refactor commands to make better use of TestSuite objects

* refactor test code into a TestSuite object and install_test module

* update mpi tests

* refactor tests suites to use content hash for name and record reproducibility info

* update unit tests and fix bugs

* Fix tests using data dir for new format
Use new `self.test_stage` object to access current data dir

Co-authored by: Tamara Dahlgren <dahlgren1@llnl.gov>

* fix setup of run environment for tests

* remove unnecessary 'None' option from run_tests status arg

* allow package files for virtuals

* run tests for all virtuals provided by each package

* add tests for mpi

* add compiler tests for virtual packages

* run compiler tests automatically like virtuals

* use working_dir instead of os.chdir

* Move knowledge of virtual-ness from spec to repo

* refactor test/cmd/clean

* update cmd/pkg tests for correctness

* Smoke tests: cmake version checks

* Simplified cmake install checks: dict-to-list

Co-authored-by: Greg Becker <becker33@llnl.gov>

* Smoke tests: emacs version checks

* Preliminary install tests for the Umpire package

* Snapshot smoke tests that grab and run examples

* Resolved openmpi example test issues for 2.0.0-4.0.3

* Use spec.satisfies; copy extra packages after install (vs. prior to install tests

* Added smoke tests for selected openmpi installed binaries

* Use which() to determine if install exe exists

* Switched onus for installer test source grab from installer to package

* Resolved (local) flake8 issues with package.py

* Use runner.name; use string format for *run_test* messages

* Renamed copy_src_to_install to cache_extra_test_source and added comments

* Metadata path cleanup: added metadata_dir property to and its use in package

* Support list of source paths to cache for install testing (with unit test)

* Added test subdir to install_test_root; changed skip_file to lambda

* Refined openmpi version checks to pass for 2.1.0 through 4.0.3

* Allow skipping install tests with exe not in bin dir and revised openmpi version tests

`spack test run --fail-first` exits after the first failed package.

`spack test run --fail-fast` stops each package test after the first
failure.

previously, tests would fail on the first error
now, we wrap them in a TestFailure object that records all failures

* spack test: subcommands for asynchronous tests

* commands are `run`, `list`, `status`, `results`, `remove`.

* tests: Preliminary libsigsegv smoke tests (updated)

* Cleaned up and added doc to libsigsegv smoke test

* tests: Update openmpi smoke tests to new run_test api

* Removed version check try-except tracking per discussion

* Changed openmpi orted command status values to list

* install tests: added support for multiple test command status values