memory_store_tests::ensure_full_copy_of_bytes_is_made_test regularly failing

[chrisstaite-menlo]

During the ASAN build the test memory_store_tests::ensure_full_copy_of_bytes_is_made_test regularly fails.

failures:

---- memory_store_tests::ensure_full_copy_of_bytes_is_made_test stdout ----
thread 'memory_store_tests::ensure_full_copy_of_bytes_is_made_test' panicked at 'Memory usage increased by 1.3929403332736068 perc, which is more than 1.2 perc', cas/store/tests/memory_store_test.rs:109:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[aaronmondal]

I keep encountering this as well during CI testing. I raised the limit from 1.1 to 1.2 already, but asan causes so much increase in memory usage that we'd probably need something like 1.6-1.8 here. I'm torn whether that's the right approach though since it would reduce the usefulness of the test for non-asan runs.

Wdyt about disabling it for the asan workflow entirely? Or we could try to make memory usage of turbo cache more predictable?

Another interesting option that I believe @allada already played around with is to use a different allocator altogether which likely has a big influence on this test. There is a good chance that tcmalloc or mimalloc could give us lower memory usage and an overall performance improvement essentially for free.

[allada]

I'd personally prefer to just disable the test for all sanatizers. This test is a very specific regression test, so I think it's ok to disable it under certain scenarios.