venus: protocol comamnd stream codec optimizations
The obvious optimizations that would end up with non-trivial overhead reduction are:
- encoder/decoder: optimize to do more assignments instead of memcpy
- where to contribute => https://gitlab.freedesktop.org/virgl/venus-protocol
- decoder: obj lookup optimization
- be ctx aware
- avoid single hashtable
- do lru for hotspot
- where to contribute => https://gitlab.freedesktop.org/virgl/virglrenderer