Speed up NewShaX

[qmuntal]

In many places we create a hash object, e.g. by calling openssl.NewSHA256(), just to get the type of hash and map it to an OpenSSL hash identifier. Then the original hash is discarded. These pattern is common enough to worth optimizing.

This PR refactors the evpHash implementation so that the hash EVP_MD_CTX context is created and initialized lazily. These are the speed gains:

goos: windows
goarch: amd64
pkg: github.com/golang-fips/openssl/v2
cpu: Intel(R) Core(TM) i7-10850H CPU @ 2.70GHz
             │    old.txt     │               new.txt               │
             │     sec/op     │   sec/op     vs base                │
NewSHA256-12   1533.50n ± 27%   50.39n ± 8%  -96.71% (p=0.000 n=10)

             │  old.txt   │            new.txt             │
             │    B/op    │    B/op     vs base            │
NewSHA256-12   24.00 ± 0%   24.00 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

             │  old.txt   │            new.txt             │
             │ allocs/op  │ allocs/op   vs base            │
NewSHA256-12   1.000 ± 0%   1.000 ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

[dagood]

Why is this KeepAlive necessary? (And why now?)

[qmuntal]

Not really necessary, as d is referenced after accessing the context. But I get nervous when I see a missing KeepAlive, as a refactor could break the code.