Improve memory buffering

[RReverser]

I'm using mmap-ed RAW files as the source of data, which should provide best performance as all the data is directly mapped into memory and accessible just &[u8].

Unfortunately, rawler can't benefit from this yet, because 1) RawFile requires only &'static data, which can be worked around by giving it a custom newtype that preserves the mmap handle, but is somewhat cumbersome and 2) rawler does a lot of tiny allocations & copying in methods like subview, which are unnecessarily expensive, as with sources like Mmap or even just Vec it could instead use subslices from existing memory.

In particular, each subview_until_eof copies almost the entire file into new and new allocations over and over. At that point it would be cheaper to have one central Vec<u8> and use &[u8] subslices everywhere.

[RReverser]

After thinking about this a bit more, I think the best solution here would be to use the common bytes crate: https://docs.rs/bytes/latest/bytes/struct.Bytes.html

It allows to wrap arbitrary buffer types, provides zero-cost shared chunks without the lifetime mess, as well as provides some useful parsing helpers.

[cytrinox]

May have a look at #416, feedback welcome.