How to reduce the IVC tests duration

[jpraynaud]

Why

We want to reduce the execution time of the IVC tests which are running for hours and could probably be optimized.

What

Assess the refactoring and optimizations that could be done to reduce the IVC tests.

How

• Profile the execution of the slow tests to identify bottlenecks
• Redefine the test strategy for the very long running tests of IVC proof system by:
  • Reducing their number
  • Reducing the redundancy of test across modules
  • Reusing pre-computed assets
  • Shard the tests (see https://github.com/input-output-hk/mithril/pull/3345/changes#diff-1cd9e2024b17ec3d7f48088edbfafde75f79366cbe5c83e01d55bc76a67667d1R127)
  • Mocking / behavior tests
  • TBD

[perturbing]

Just to log it for next week.

Base run against main: 1h 15m 18s

Reuse SRS by writing to disk for tests run (given this commit): 1h 1m 6s

shard the snark tests on this run with this commit: 32m 35s

Note that currently we still can test if the opt-level=3 is also worth it in the sharded setting (run time is now ~15m for the actual tests, depends on the shard)

Also, the current sharding technique build the tests witouth sharing the builds, this can be cached via cargo nextest archive