add multi-block read for performance

[yuiiio]

code generated by claude

Although the code was generated by AI,
it has been formatted to be sufficiently compact and readable (including alignment adjustments),
and does not interfere with other code.

---

perf improvement
at esp32s2 & ~380KB image

current read impl: 357ms
read_multi<64 blocks at once>: 195ms

The stack size must be sufficient to accommodate the number of blocks at onece.

[yuiiio]

I think this will take over from #46