Update + Lossy Compression

[jnsiemer]

Description

This PR does the following:

• ensure that its implementation is up-to-date with the current version of math and uses the correct version
• adds lossy compression as defined for ML-KEM with implementations for PolynomialRingZq and MatPolynomialRingZq.

[Marvin-Beckmann]

I would make the following suggestions:

1. make the output of the compression have the different explicit modulus as described in the FIPS document. Therefore I suggest to either: Make a new modulus object, that you potentially provide, or: make the output type only capture the values of the different coefficients/coordinates of coefficients. (Otherwise inintended behavior can easily occur if the compressed values are used beyond just compressing and decompressing
2. I think individual functions rather than a trait is a bit better here personally. Especially if you implement the previous suggestions, a trait does not necessarily make too much sense anymore. Also: the traitname suggests that by implementing it like this: this lossycompression is the only possible lossycompression technique there is. If you make the name more tailored to the actual functionality, then you can also add individual function names, which is what I would opt for here.

[Marvin-Beckmann]

same comments