Command to access logs systematically without breaking .stack-work's abstraction

There are several issues here that discuss which output should and should not be shown under various conditions. But in all cases the fallback is to look at the logs.

Currently, we just dig around in `./.stack-work/logs` and make assumptions about its layout. Is that the intended method?

Should there instead be a way to run `stack logs ` to print out specific logs in a more future proof way?

P.S. This is especially important for benchmarks, which are currently run ONLY for their logs .

Start by examining the current assumptions around `./.stack-work/logs` and how benchmarks are run using their logs. Compare the requested `stack logs ` entry point with the existing output behavior, then clarify which specific logs and conditions the command must support before implementing it.

Written by the indexing model from the issue text.