Add tests to check data CRC

[jongiddy]

The code currently has no tests for the data CRC. Add a test that the GzEncoders generate the CRC expected by RFC 1952 and that the GzDecoders fail if the CRC does not match.

This will give us more confidence to make any changes to the CRC calculation (e.g. #523 or #526).

[Byron]

Thanks a lot, I also think these will be very useful!

I'd wave the PR through, but then again, I have no clue. Maybe @folkertdev could take a look to see what he thinks might be missing. Based on this, we can decide if there should be more, or of that 'more' can or should be in another PR.
But maybe there are legitimate change requests as well, so I'd want to wait for his opinion.

[folkertdev]

Some minor style ideas/suggestions. Overall I think this gives good coverage of how flate2 uses checksums.

my PR additionally tests the public API (e.g. combine and reset). So together I think that would cover crc usage well.

[Byron]

Thanks a lot for chiming in @folkertdev! @jongiddy, I will wait until you had a chance to look at the suggestions, but generally will be happy to merge.

[jongiddy]

Thanks for the great suggestions. I have applied them all.

[Byron]

Thanks a lot everyone! I'd merge this.

However, please allow one last question: could these tests be in tests/ as they don't use private APIs? It looks like it, and I would have checked (and possibly done) the move myself but then I can't merge it anymore 😅.

Then again, as this doesn't introduce tests, we could also move the tests over, if possible, in a separate PR and call this one done enough.

[jongiddy]

The tricky part of moving to tests is that the new tests uses the Rfc1952Crc type, which is also used by the roundtrip_header test which does use private APIs (GzBuilder::into_header).

[Byron]

Ah, I see, thanks for clarifying.