chore(ci): improving CI

[jeertmans]

Hello @maciejhirsz,

This PR proposes new changes regarding continuous integration tests, that I think should facilitate maintaining this project.

Here are the main advantages:

• improved speed for workflows with cache intensive usage; and a specific workflow is used to clear cache on closed PR (clearcache.yml);
• code coverage (tests), with badge (e.g.: see my README) and PR reports (e.g.: test(ci): add dumb token jeertmans/logos#2 (comment)). Setup is very easy: https://app.codecov.io/;
• benchmarking PR before/after changes (e.g.: test(ci): add dumb token jeertmans/logos#2 (comment));
• test Rust minimal support version (MSRV). In my opinion, this is always good to document the MSRV for your project, and a workflows automates it;
• tests are split in multiple workflows and jobs, so that contributors easily understand what fails;
• and many more tests are provided.

A few remarks:

• I update Rust's version and Rust's edition;
• benchmarks are now performed using Criterion (it automatically compares benchmarks);
• the benchmark action should fail here, this is because the master base branch does not contain the expected benchmarks; it will work after merging;
• you need to setup https://app.codecov.io/ for the code coverage upload to work;
• and Clippy's check should fail, because it already fails on the base branch. But I think fixing its warning should be for another PR.

Do no hesitate to give your opinion on this :-)

[maciejhirsz]

Thanks @jeertmans! I'll have a proper look at this hopefully later today.

I'm trying to manage my time a bit better, and I'm thinking about carving out one day a week for a general maintenance sweet across my crates, tentatively looking at Fridays. I reckon the pace being slow here will be much more manageable if it's at least predictable.

[jeertmans]

No problem @maciejhirsz
I hope this PR will also facilitate reviewing future PRS :D

[jeertmans]

Results for before/after changes on applying clippy's suggestions

group                                         before                                 changes
-----                                         ------                                 -------
count_ok/identifiers                          1.00   488.3±20.25ns  1521.4 MB/sec    1.01   495.0±15.08ns  1500.9 MB/sec
count_ok/keywords_operators_and_punctators    1.01  1366.1±11.08ns  1487.6 MB/sec    1.00  1350.6±42.38ns  1504.7 MB/sec
count_ok/strings                              1.00   536.3±10.77ns  1548.8 MB/sec    1.00   537.9±18.65ns  1544.4 MB/sec
iterate/identifiers                           1.00   485.4±13.38ns  1530.5 MB/sec    1.05   509.0±19.42ns  1459.7 MB/sec
iterate/keywords_operators_and_punctators     1.00  1342.7±41.67ns  1513.6 MB/sec    1.01  1356.0±32.37ns  1498.8 MB/sec
iterate/strings                               1.00   545.3±15.31ns  1523.3 MB/sec    1.00    542.9±7.48ns  1530.1 MB/sec

Conclusion

It did not affect performances (as expected).

[jeertmans]

@jeertmans, I'm merging this in the hope it will accelerate developing the book and the "new" Logos, thanks to a better test-suite.

If you are okay with test coverage, you still need to setup (free) the codecov app:

• code coverage (tests), with badge (e.g.: see my README) and PR reports (e.g.: test(ci): add dumb token jeertmans/logos#2 (comment)). Setup is very easy: https://app.codecov.io/;

Once done, we can add a badge on the README.