Faster update twiddles function in whir crate

[vladfdp]

Did the TODO in [dft.rs](http://dft.rs) of the whir crate.
instead of recomputing everything we reuse the existing twiddles.
Here are benchmarks run in release mode comparing the new function to the old one.
The speedup is significant but drops down with the ratio of new to old twiddles size

start        target       update_twiddles     update_new    speedup
2^8          2^9                 1.816µs          537ns      3.38x (2x growth)
2^8          2^10                3.062µs        1.414µs      2.17x (4x growth)
2^8          2^11                7.926µs        3.295µs      2.41x (8x growth)
2^8          2^12               14.193µs        7.358µs      1.93x (16x growth)
2^12         2^13               25.152µs        7.051µs      3.57x (2x growth)
2^12         2^14               46.344µs       15.242µs      3.04x (4x growth)
2^12         2^15               86.562µs       35.265µs      2.45x (8x growth)
2^12         2^16              162.497µs       90.329µs      1.80x (16x growth)
2^16         2^17              328.373µs        97.45µs      3.37x (2x growth)
2^16         2^18              653.316µs      218.411µs      2.99x (4x growth)
2^16         2^19             1.298278ms      544.642µs      2.38x (8x growth)
2^16         2^20             2.603312ms     1.416323ms      1.84x (16x growth)

[EXDEV-oss]

Hey bro, I came across your project on GitHub and I genuinely like the direction you’re taking with it, especially the backend/infrastructure side.

I normally work around testing, project exposure, and connecting solid builders with small but effective growth networks.

Not here to waste your time or sell anything weird lol — I just like connecting with serious devs early before projects get bigger.

Are you currently building this for personal scaling, startup plans, or long-term launch?