Implement one shot SHA and reduce allocs

[qmuntal]

This PR contains two performance improvements which landed into dev.boringbranch during this development cycle, both committed in https://go-review.googlesource.com/c/go/+/395876/11.

• a99ec67c99c0e0d2bc69e495dbb0794969ec3314: Implement one shot SHA calls. This is helpful for callers which don't need to keep the hash object returned by NewSha1 and friends, as these functions return a pointer which normally escapes in the caller side.
• 9fafecd26c83a05018bd5a64541c488c2cb247db and 31857f33f60aef96df3b40722c163ebdf13b6fc5: Makes SHA calls allocation-free by hiding openssl pointers from the escape analysis.

The benchmarks demonstrates that the new code does not allocate:

name          old time/op    new time/op    delta
Hash8Bytes-4    4.57µs ± 4%    4.46µs ± 4%    -2.55%  (p=0.009 n=10+10)

name          old speed      new speed      delta
Hash8Bytes-4   224MB/s ± 4%   230MB/s ± 4%    +2.62%  (p=0.009 n=10+10)

name          old bytes/op   new bytes/op   delta
Hash8Bytes-4     24.0B ± 0%      0.0B       -100.00%  (p=0.000 n=10+10)

name          old allocs/op  new allocs/op  delta
Hash8Bytes-4      1.00 ± 0%      0.00       -100.00%  (p=0.000 n=10+10)

Removing the SHA allocations is important for when openssl is integrated into go1.19, else we would have to disable a new test that checks boring SHAs don't allocate.

[qmuntal]

I've found something good: noescapeCtx is not necessary because we are passing OpenSSL C pointers as uintptr instead of unsafe.Pointer, and the former are already safely hidden from the escape analysis.

So there are only three remaining usages of noescape, but we can't avoid them because they work on Go pointers.