dev: processor optimizations

[ldez]

I was working on the new exclusion configuration #5297, and I found some "optimizations".

• IdentifierMarker processor update:
  • removes unused regexp
  • checks the match only if the issue is from a linter with the associated pattern instead of trying to match all regex on all issue texts.
  • adds one test by pattern.
  • Note: I'm unsure if keeping/removing this processor in the future is a good idea.
• new PathAbsoluter processor:
  • the absolute paths were obtained by the Cgo processor and the FilenameUnadjuster then now the PathAbsoluter does that once.
  • in theory, all the paths from analysis.Diagnostic are absolute.
  • FYI the processor PathPrettifier (it already exists) converts absolute path to relative (related to the different elements but mainly to exclusions)

This is not only "optimizations", the PathAbsoluter is something I will use with the new exclusion system.

I found another optimization: the usage of FileCache inside misspell.
The FileCache is used by SourceCode, ExcludesRules, severity, Fixer processors, and misspell.
The FileCache is loaded with file content and the keys are paths, but misspell is the only one that uses absolute paths to load/read data from/to the cache.
So the cache is filled but never used.
This allows to reduce the memory usage (ex: on k8s this 50MiB)

These are minor optimizations: a few seconds, but they can impact large projects.

To avoid creating a PR for exclusions with some off-topics, and as those elements are standalone, I prefer to create a PR with just these processors.

[bombsimon]

Nice!

Re IdentifierMarker, I think it's a bit weird. It feels like it's from early days where the linters and their diagnostics was more controlled. Surely I'd like to have identifiers within backtiks just as much as the next person, but having a processor doing search replace on hand written regexes for each linter feels a bit unnecessary and hard to maintain. I'm sure there are more identifiers out there we don't replace, or replacements that are removed from the upstream linters anyways.

[ldez]

I agree, we will not remove it for now, I want to avoid "breaking changes" but we will not add new regexp unless there is a real need.