Add quantization codec and implement compression/decompression of floating point data

[astrofrog]

Fixes #14
Fixes #40

[github-actions]

👋 Thank you for your draft pull request! Do you know that you can use [ci skip] or [skip ci] in your commit messages to skip running continuous integration tests until you are ready?

[astrofrog]

I think lossy tests are close to passing but due to the issue mentioned in #40 the floating point values don't match exactly:

E       Mismatched elements: 144 / 144 (100%)
E       Max absolute difference: 7.677353
E       Max relative difference: 7.3442245
E        x: array([[163.9622  ,  12.696566,   2.805807,  83.07929 ,   3.645739,
E                12.367053, 118.553955,  33.24904 ,  59.632137, 209.43936 ,
E                23.05459 ,  67.77326 ],...
E        y: array([[1.626797e+02, 9.415493e+00, 1.536737e+00, 8.256937e+01,
E               5.992008e+00, 9.394189e+00, 1.165274e+02, 3.576727e+01,
E               6.091895e+01, 2.120618e+02, 2.193155e+01, 6.678799e+01],...

I think we should adjust the tests to compare the data read in by both libraries rather than comparing to the very original data.

[Cadair]

I don't think we can have this inherit Codec if we aren't going to implement encode and decode as it will fail the ABC check if numcodecs is importable.

[astrofrog]

That's fine and we are going to remove references to Codec anyway so we can just change this to not inherit from anything?

[Cadair]

This might want a better variable name, you can have lossless data in the COMPRESSED_DATA column if it was compressed directly with GZIP without quantization? Am I right in thinking that this is "this tile is losslessly compressed when the others are quantized"?

[astrofrog]

Yes that's correct

[astrofrog]

gzip_fallback?

[astrofrog]

@Cadair - I found some time to implement the comments