perf: adds AVX512 implementations of vector.Sum, vector.InnerProduct …

…+ assembly refactor (#547)

* feat: asm Vector sum slower, no avx

* checkpoint

* checkpoint

* checkpoint

* feat: add vec.Sum AVX512

* test: make odd bound for better test case

* build: make linter happy

* fix: update bound for vec sum to match parameter choices

* perf: loop 8 by 8, cosmetics

* style: cosmetics

* test: better sum test

* test: more test

* refactor: move common assembly routine in subfolder (#545)

* refactor: move common assembly routines in root

* build: make linter happier

* style: cosmetics

* test: start fixing integration test

* style: factorize mul documentation

* feat: add .ASMVector and fix integartion test

* test: fix 32bit test

* test: fix previous commit

* doc: add reference for reduction algorithm

* feat: use latest bavard for avx512 instructions

* feat: added purego InnerProduct

* checkpoint wip

* checkpoint

* refactor: checkpoint

* test: better tests for vec ops

* checkpoint

* test: add more tests for vector ops

* feat: update bavard and use better syntax in asm

* test: make benchmarks on varying sizes

* test: bench on larger vector

* test: bench on larger vector

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* checkpoint

* refactor: use defines for mul

* feat: make use of defines in assembly

* checkpoint

* style: code cleaning

* perf: prefetches in vec ops

* perf: minor adjustements

* style: costmetics

* feat: handle case where len(vec)==0

* fix: address PR review comments