perf(prover): reduce synchronization allocs for FFT

[estensen]

Replace channel-based synchronization with sync.WaitGroup.

I think the primary reason is eliminating channel allocs in the recursive calls.

Benchmark results on M4 Pro show ~17% reduction in memory allocations (from ~8.2KB to ~6.7KB per operation) while ns/op remains similar.

Before

pkg: github.com/consensys/linea-monorepo/prover/maths/fft
BenchmarkFFT/fft_2**8-12                  127771              9392 ns/op               0 B/op          0 allocs/op
BenchmarkFFT/fft_2**9-12                   41942             28490 ns/op            1128 B/op         19 allocs/op
BenchmarkFFT/fft_2**10-12                  29746             40334 ns/op            2616 B/op         45 allocs/op
BenchmarkFFT/fft_2**11-12                  17817             67836 ns/op            4824 B/op         85 allocs/op
BenchmarkFFT/fft_2**12-12                  10000            111118 ns/op            8217 B/op        149 allocs/op
BenchmarkFFT/fft_2**13-12                   6920            174276 ns/op            8216 B/op        149 allocs/op
BenchmarkFFT/fft_2**14-12                   4262            279231 ns/op            8216 B/op        149 allocs/op
BenchmarkFFT/fft_2**15-12                   2568            492729 ns/op            8216 B/op        149 allocs/op
BenchmarkFFT/fft_2**16-12                   1305            890350 ns/op            8216 B/op        149 allocs/op
BenchmarkFFT/fft_2**17-12                    686           1762311 ns/op            8216 B/op        149 allocs/op
BenchmarkFFT/fft_2**18-12                    333           3620180 ns/op            8218 B/op        149 allocs/op
BenchmarkFFT/fft_2**19-12                    157           7404848 ns/op            8216 B/op        149 allocs/op

After

BenchmarkFFT/fft_2**8-12                  126274              9503 ns/op               0 B/op          0 allocs/op
BenchmarkFFT/fft_2**9-12                   41919             28420 ns/op            1032 B/op         19 allocs/op
BenchmarkFFT/fft_2**10-12                  29798             40198 ns/op            2328 B/op         45 allocs/op
BenchmarkFFT/fft_2**11-12                  17816             67560 ns/op            4152 B/op         85 allocs/op
BenchmarkFFT/fft_2**12-12                  10000            111115 ns/op            6777 B/op        149 allocs/op
BenchmarkFFT/fft_2**13-12                   6847            182320 ns/op            6776 B/op        149 allocs/op
BenchmarkFFT/fft_2**14-12                   4046            277555 ns/op            6776 B/op        149 allocs/op
BenchmarkFFT/fft_2**15-12                   2565            463729 ns/op            6776 B/op        149 allocs/op
BenchmarkFFT/fft_2**16-12                   1388            876708 ns/op            6777 B/op        149 allocs/op
BenchmarkFFT/fft_2**17-12                    682           1774575 ns/op            6778 B/op        149 allocs/op
BenchmarkFFT/fft_2**18-12                    336           3605445 ns/op            6776 B/op        149 allocs/op
BenchmarkFFT/fft_2**19-12                    160           7456658 ns/op            6776 B/op        149 allocs/op

[cla-assistant]

All committers have signed the CLA.