Codex api/disc ports were clashing in windows because `lastUsedCodexApiPort` and `lastUsedCodeexDiscPort` were not locked when read and written, while client and provider nodes were started concurrently (async).
DEBUG -> enables DebugCodexNodes, DebugTestHarness, and ShowContinuousStatusUpdates
DEBUG_HARDHAT -> enables DebugHardhat
PARALLEL -> enables EnableParallelTests
Additionally, when DEBUG is enabled, all integration tests debug configs are enabled for Codex nodes, the Codex node output is printed with the test output (not interleaved), and the Codex node output is logged to file in `tests/integrations/logs/<starttime>__IntegrationTests/<integration_test_name>/<suite_name>/<testname>/<role>_<idx>.log`.
When DEBUG_HARDHAT is enabled, all hardhat output is printed with the test output (not interleaved), and the output is also written to a log file in `tests/integrations/logs/<starttime>__IntegrationTests/<integration_test_name>/hardhat.log
Previously, each test would start with a port specified by a global const and overwritable using `define` cli flag. This seems to be causing some port clashes that `nextFreePort` is not picking up. Instead of relying on `nextFreePort`, do not reuse port numbers for any of the tests.
Solution was two-fold:
1. Run the hardhat.cmd script instead of hardhat
2. Use the absolute path to the script executable instead of relative to the working directory
- bubble errors from hardhatprocess.start (instead of just logging)
- push raises: [] in all nodeprocess procs/methods to avoid leaking `Exception` exception types in method overrides
- add a TestManager property to IntegrationTest, so manager does not need to be passed into all functions
- cleanup:
- remove unneeded stopHardhat function
- add hardhat instance to manager outside of startHardhat
