Speed up from_bitmap_bytes, and use less unsafe

[Dr-Emann]

• Directly create an array/bitmap store based on the count of bits rather than always making a bitmap, and converting to the correct store after
• Use u64 words to count bits and enumerate bits (this appears to make a pretty big difference)
• Store in little endian, then just fixup any touched words
• Only use unsafe to reinterpret a whole bitmap container as bytes, everything else is safe
• Allow adding bytes up to the last possible offset

On my machine (an m1 mac), this speeds up the benchmarks by between 16%-120% (probably based on how sparse the data is)

See RoaringBitmap#288

[lemolatoon]

@Dr-Emann Thank you for sending patch! I have just reviewed your code.

• It is surprising that using u64::count_ones speeds up so much. Nice.
• Defining from_lsb0_bytes for each store makes much sense.
• Converting endians after writing bytes seems to be good, making code clearer.

I'll merge this.