deflate/zip: impl compress_stream() and optimizations

[ManuLinares]

this was a chore to merge
I'll review this later

implement `compress_stream`

dyn huffman + compress fastpass

various optimizations

11-bit lookup table for Huffman decoding

128bit simd vector moves for matches to maximise throughput

rebuilt the bitstream reader around a 64bit buffer and refactored the
decoder into a tight hot loop to eliminate statr machine overhead and
align the sliding window to 64bit boundaries

add `test_deflate_embedded_stream()` test

add more zip tests

[data-man]

I think that StreamBitReader and BitWriter can be useful as public in std::io.
What do you think?

[lerno]

Please look at deflate, because I had to update it.

[ManuLinares]

Please look at deflate, because I had to update it.

Done, I still couldn't debug why it breaks in windows and alpine.

[lerno]

First thing I'd check is if it works if the vector usage is removed.