Add LRU caches for performance improvements

[aiooss-ledger]

A first step towards https://github.com/Ledger-Donjon/rainbow/issues/6

I am using a TVLA script on a firmware running Rust aes crate to benchmark:

• Before: 4.34 traces/s
• Add LRU cache on Capstone calls: 7.54 traces/s
• Add LRU cache to get written registers: 11.12 traces/s
• Import regs_hw_sum_trace outside of sca_code_trace: 12.56 traces/s