AVX2 fnvHash. Bump numOfGoroutines to 8 in stage_senders.go (#656)

* Profile all stages

* Try to recover senders with 8 goroutines

* fix CPU profiling for stage_bodies

* fix out-of-index

* Try full DAG for verfication of header seals

* Try to unroll fnvHash for performance

* SSE2 assembly for fnvHash16

* fnvHash16AVX2

* Revert changes to state.go

* check we're on 64-bit in useAVX2

* Shave a move off fnvHash16AVX2

* asmdecl doesn't know about VMOVD

* disable linter in the right place