Compression changes to eo-learn

[mlubej]

Changes:

• compression_level parameter removed
• compress by default for everything except bbox and timestamps
• test updates

[mlubej]

@zigaLuksic turns out that backwards compatibility already worked out of the box, since the featureIO objects retained the ability to handled loading of compressed/uncompressed objects

I added back the compression_level parameter as discussed, and added the EODeprecation warning on the save task and the save method. I removed the parameter from the docs, though.

Lastly, I added tests for checking the compression deprecation warning, and augmented a test for checking backwards compatibility loading and proper cleanup if different compressions are present (from before)

[zigaLuksic]

@zigaLuksic turns out that backwards compatibility already worked out of the box, since the featureIO objects retained the ability to handled loading of compressed/uncompressed objects

Slight correction, the FeatureIO objects retained the GZip capabilities so that we have backwards compatibility ;)
After a while, when this is removed, we'll be able simplify FeatureIO classes

[mlubej]

Also, I wasn't sure if I should keep the compression capabilities in the save_eopatch function (from eolearn/core/eodata_io.py). This one should be internal, so people shouldn't be using it, and the tests might've looked a bit better, but dunno, someone still might use it.

[zigaLuksic]

Also, I wasn't sure if I should keep the compression capabilities in the save_eopatch function (from eolearn/core/eodata_io.py). This one should be internal, so people shouldn't be using it, and the tests might've looked a bit better, but dunno, someone still might use it.

I see no reason to keep it, since it would control only part of the compression behaviour (for numpy and geodata) which we dont really want to be changed (except for tests)