Add HHClientLinter

[Atry]

This is the initial implementation of HHClientLinter, which is a proxy calling hh_client --lint.

[fredemmott]

• test failure: either use a lint for the test that exists in all supported versions, or increase the minimum version (grep for the old one, probably composer.json and .github/workflows/*)
• needs adding to LintRunConfig::NON_DEFAULT_LINTERS so it's include in all runs.
  • this will make it so that it's include in HHAST's own runs; probably best to exclude this linter explicitly in hhast-lint.json for now, then fix them up separately

[fredemmott]

The active lints need to be configurable; perhaps:

shape(
  ?'ignore_except' => keyset<int>,
  ?'ignore' => keyset<int>,
 )

Both would default to empty; if something is present in both, that should be a runtime error.

Additionally, there should probably be a hardcoded list of lints to always ignore, e.g. 5583 - await in a loop - is known buggy, and has a pure Hack implementation in HHAST and Facebook www already. The current list is HackLintRule::HACK_LINT_IGNORED_CODES in FB WWW, mostly being known-bad linters that should be fixed at some point, but for 5583/AWAIT_IN_LOOP, this is permanently replaced by the linters written in Hack so blocking it here (and in FB) is a band aid - the right fix is probably to delete that one from the ocaml code.

[Atry]

I will implement the config in another PR as this PR is already too large

[Atry]

• test failure: either use a lint for the test that exists in all supported versions, or increase the minimum version (grep for the old one, probably composer.json and .github/workflows/*)

Fixed

• needs adding to LintRunConfig::NON_DEFAULT_LINTERS so it's include in all runs.

  • this will make it so that it's include in HHAST's own runs; probably best to exclude this linter explicitly in hhast-lint.json for now, then fix them up separately

Will be addressed in another PR

[fredemmott]

note followup PRs (to add to non-default linters, and the configurability) are required before this can actually be used outside of unit tests.

[fredemmott]

this should probably be memoized or passed in to the constructor given that it's never going to change for a particular instance

[Atry]

By the way, Hotspot JVM is good at optimizing allocation for short-lived objects like this (when it is not memoized). Is HHVM able to eliminate the allocation like JVM does?

[fredemmott]

Sorry, I'm not familiar enough with either VM:)