perf(ws): optimize fastwebsockets in release profile

[lucab]

This sorts and re-aligns the bench and release profiles, notably enabling optimizations for fastwebsockets in the released CLI binary.
The goal is to unlock auto-vectorization for ws hotpaths, as noted in https://github.com/denoland/fastwebsockets/blob/0.7.0/src/mask.rs#L22 (contrarily to that comment, disassembling the amd64 deno 1.44 binary shows no SIMD instructions in ws-unmasking logic).

[littledivy]

(contrarily to that comment, disassembling the amd64 deno 1.44 binary shows no SIMD instructions in ws-unmasking logic)

oh interesting 🤔 we should probably re-enable explicit SIMD then

[littledivy]

LGTM

[lucab]

we should probably re-enable explicit SIMD then

I think you don't have to.
On my local 1.79 rustc, auto-vectorization on amd64 seems to kick in at opt-level >= 2.
And after this PR, on the latest canary from CI, I can indeed see that the ws hotpath is now using SIMD instructions for unmasking:

00000000056ecc10 <_ZN14fastwebsockets5frame5Frame6unmask17h1730290b8ecf1d1cE>:
 56ecc10:       55                      push   %rbp
 56ecc11:       41 57                   push   %r15
[...]
 56ecd78:       49 ff c1                inc    %r9
 56ecd7b:       4d 21 cd                and    %r9,%r13
 56ecd7e:       4a 8d 0c ae             lea    (%rsi,%r13,4),%rcx
 56ecd82:       66 41 0f 6e c0          movd   %r8d,%xmm0
 56ecd87:       66 0f 70 c0 00          pshufd $0x0,%xmm0,%xmm0
 56ecd8c:       45 31 d2                xor    %r10d,%r10d
 56ecd8f:       90                      nop
 56ecd90:       f3 42 0f 6f 0c 96       movdqu (%rsi,%r10,4),%xmm1
 56ecd96:       f3 42 0f 6f 54 96 10    movdqu 0x10(%rsi,%r10,4),%xmm2
 56ecd9d:       66 0f ef c8             pxor   %xmm0,%xmm1
 56ecda1:       66 0f ef d0             pxor   %xmm0,%xmm2
 56ecda5:       f3 42 0f 7f 0c 96       movdqu %xmm1,(%rsi,%r10,4)
 56ecdab:       f3 42 0f 7f 54 96 10    movdqu %xmm2,0x10(%rsi,%r10,4)
 56ecdb2:       49 83 c2 08             add    $0x8,%r10
 56ecdb6:       4d 39 d5                cmp    %r10,%r13
 56ecdb9:       75 d5                   jne    56ecd90 <_ZN14fastwebsockets5frame5Frame6unmask17h1730290b8ecf1d1cE+0x180>
[...]