fix: api itest panic

[aarshkshah1992]

1. Fixes flakiness in testNonGenesisMiner in api_test by de-racing the closing of the miner repo and the sealer reading it in a go-routine. If the go-routine gets scheduled after the miner repo is closed in the test, this causes a panic which fails the test. The right thing to do here is for the sealer to read the repo when it is instantiated and pass the config to the go-routine.

2. Adds some logging to the testOutOfGasError itest to see what's going on when it fails (very very rare).

[rvagg]

This breaks 2 public APIs, which while we may not expect people to be using them, we should at least signal and not just roll this out in a casual patch release.

My preference for signalling breakage is in commit message, which, if we can do consistently, will really help our changelog, so your commit message could be something like:

fix!: sealer: handle initialisation error without panic

storage/pipeline.NewPreCommitBatcher and storage/pipeline.New now have an additional
error return to deal with errors arising from fetching the sealing config

Also API breakage may be something we shouldn't skip changelog on?

[rjan90]

Also API breakage may be something we shouldn't skip changelog on?

Yes, we should add this the changelog / write down what the API´s we are breaking - so its clear/understandable what should be added in a Upgrade Warnings section of a new release.

[aarshkshah1992]

@rvagg Addressed your review. Please can you review ?

[rvagg]

for the record, examples of failures that we're attempting to fix here:

• https://github.com/filecoin-project/lotus/actions/runs/9891743740/job/27322860947
• https://github.com/filecoin-project/lotus/actions/runs/9891029052/job/27320620791
• https://github.com/filecoin-project/lotus/actions/runs/9949190437/job/27485039787