Adjust cache for cargo

[citizen-stig]

Short summary

• Add Swatinem/rust-cache@v2 before sccache, so we can combine crates/target cache with recompilation.
• Adjust macro tests to not require native flag
• change assert!(matches!()) to assert_eq!()
• Switch to use latest sccache instead of hard coded.

Description

Currently we use sccache for caching compilation. and it helps, especially with rocksdb, as it can cache not only rust compilation artifacts, but also C/C++.

But it looks like cargo download dependencies everytime and scacche does not handle it that well.

So idea is to put sccache on top of standard rust caching mechanism.

Iteration	Check	Test	Coverage	Features	Check Prover
1 (empty cache)	13m	23m	23m	30m	45m
2	9 m	19m	20m	23m	22m
3 (add cache for target)	12m	26m	20m	25m	24m
4	2m	7m	20m	3m	10m
5	1m	9m	10m	4m	14m
6 (adjust macro test)	12m	15m	17m	28m	16m
7	1m	9m	7m	3m	10m

Macro tests

Current macro tests use native flag for testing compilation, which brings compilation of rocksdb second time inside macro test.
I've changed the tests to operate on ZK conteext and ZK storage, as it require less dependencies.

Future improvements

Things that can be done to make CI even smoother:

1. Set check step to be the first and rest depend on it, so it doesn't waste runner's time if it fails
2. Figure out why llvm-cov spends 10 minutes recompiling.
3. tune SCCACHE_CACHE_SIZE to get best results
4. Try cargo-nextest

[codecov]

Codecov Report

Merging #439 (d117eef) into main (524dd2a) will decrease coverage by 0.2%.
The diff coverage is 98.2%.

❗ Current head d117eef differs from pull request most recent head e8f2f1e. Consider uploading reports for the commit e8f2f1e to get more accurate results

Impacted Files	Coverage Δ
...implementations/sov-sequencer-registry/src/call.rs	94.2% <94.2%> (ø)
...m/module-implementations/sov-accounts/src/hooks.rs	92.5% <100.0%> (ø)
...lementations/sov-sequencer-registry/src/genesis.rs	100.0% <100.0%> (ø)
...mplementations/sov-sequencer-registry/src/hooks.rs	100.0% <100.0%> (+6.2%)	⬆️
...-implementations/sov-sequencer-registry/src/lib.rs	97.9% <100.0%> (+9.0%)	⬆️
...mplementations/sov-sequencer-registry/src/query.rs	75.0% <100.0%> (-7.7%)	⬇️
...ystem/sov-modules-stf-template/src/app_template.rs	82.6% <100.0%> (+1.0%)	⬆️
module-system/sov-modules-stf-template/src/lib.rs	97.4% <100.0%> (ø)

... and 3 files with indirect coverage changes

[neysofu]

The benchmarks are very impressive and all changes make sense to me. It would be nice to include a timeout (1h?) for our GH jobs, to save compute in case the CI gets stuck (it doesn't look like it happens at all by glancing at our Actions history but it's a good fail-safe to have, especially since we're doing proptests).

[citizen-stig]

The benchmarks are very impressive and all changes make sense to me. It would be nice to include a timeout (1h?) for our GH jobs, to save compute in case the CI gets stuck (it doesn't look like it happens at all by glancing at our Actions history but it's a good fail-safe to have, especially since we're doing proptests).

Agree, done!