Support readable aliases for HHClientLinter rules

[fredemmott]

#409

E.g. if we were removing DontAwaitInAloop linter and keeping 5583 (we should do the opposite, #408), we could make it so that suppressing DontAwaitInALoop suppresses 5583

• avoids the BC break
• more readable

This would be a divergence from HHAST.

This could also be more general, not just for HHClientLinter - then we could support the same suppression codes within FB with Aurora/HackAST, and externally with HHAST, when the linters do not have the same suppression code at present.

[Atry]

I think we don't want to do this manually because the aliases would be out of sync quickly when the upstream code changes. I think the best way to solve this issue is to export the metadata of OCaml lint rules to Hack automatically.

[fredemmott]

I think we need a manual aliasing method regardless, as the names in ocaml do not necessarily match the historical hhast linter names - requiring renames removes the BC b enefits.

[Atry]

I agree we should allow some overridden aliases configured in the HHAST code base for historical hhast linter names, but shall we configure a default name in the OCaml code base then generate the corresponding Hack definition, so that we could have readable aliases for new OCaml linters by default?

[fredemmott]

Though, if exported from ocaml, should probably default to the ocaml class name rather than needing explicit configuration