linfa-preprocessing::countgrams: RefCell is replaced with OnceLock

[kakserpom]

Simple fix to allow for concurrent access.

[relf]

OnceLock is not serde. What about cloning to use multi-threading?

[kakserpom]

@relf I don't quite understand what you mean by that. OnceLock is a better fit here than RefCell because it's thread-safe.

[relf]

With OnceLock we cannot serialize/deserialize CountVectorizerValidParams which is a no go (cf. CI tests errors).

I meant you can just clone CountVectorizerParams and use an independent copy in each thread (if it was your initial problem with the multi-threading).

[kakserpom]

@relf Well, I guess we could solve the issue by implementing serialize and deserialize manually, right?

[relf]

Maybe... But there is no problem in the first place if you just clone CountVectorizerParams so no need of a special serde code complication.

[kakserpom]

@relf please review the updated version.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 36.17%. Comparing base (5272ad1) to head (0a23fc7). Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #387      +/-   ##
==========================================
+ Coverage   35.90%   36.17%   +0.26%     
==========================================
  Files          97      101       +4     
  Lines        6409     6563     +154     
==========================================
+ Hits         2301     2374      +73     
- Misses       4108     4189      +81     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

• :snowflake: Test Analytics: Detect flaky tests, report on failures, and find test suite problems.