Fixes in corpus minimizer

[ThomasTNO]

Description

In an attempt to implement the corpus minimizer in TNO-S3/WuppieFuzz#146 I stumbled on some issues with the current implementation. My feeling is that parts of the code were unmaintained for too long.

I made a couple of adjustments to get it operational again.

1. Ensure inputs are loaded. Previously it would fail due to, I assume, lazy loading
2. Ensure exec_time is set. In my use case exec_time was not set as we want to minimize before starting a fuzzing run.
3. Removal of corpus entries fails sometimes. Not entirely sure why it happens, but sometimes the read ctr String contains null bytes (string has the form \u{3}000 and therefore can't be cast to u32.

For all these adjustments I am looking forward to your suggestions. In particular,

2. Ideally I would say we reuse the calibration code to calibrate the entries and ensure all required metadata is present. I could not come up with a better solution.
3. Currently the code uses read instead of read_exact which is not preferred and not allowed by the rules set in the precommit, suggestions are much appreciated. Weirdly sometimes the read buffer contains a single digit, and sometimes it contains four bytes (single digit + three null bytes). I can't get my head around it.

Checklist

• I have run ./scripts/precommit.sh and addressed all comments

[ThomasTNO]

I see 3. was already resolved in 7f0afe8

[tokatoka]

can you resolve the conflicts?

[tokatoka]

Ideally I would say we reuse the calibration code to calibrate the entries and ensure all required metadata is present. I could not come up with a better solution.

yes it's not very good.
can you maybe move the duplicate code in a public function and call them from both calibration and this minimizer code?

[ThomasTNO]

Some jobs fail due to connection errors apparently.

[tokatoka]

thank you!