Add comparisons for fxamacker/cbor, encoding/gob, encoding/json

[x448]

Benchmark comparisons to Go standard library codecs are missing.

Let's add benchmark comparisons of fxamacker/cbor, encoding/gob, and encoding/json.

Encoded size comparisons are important for some use cases. Use simple table if there isn't enough time for charts. Make it readable on phone (I wish I had this available on my phone yesterday).

Use a common data struct with realistic values for comparison.

UPDATE: something like this...

Codec	Benchmark	Duration ns/op	Memory B/op	Allocs    allocs/op	Encoded Output
fxamacker/cbor 2.5	Marshal (K,P)	201.4	176	2	80 bytes
fxamacker/cbor 2.5	Marshal (K)	421.7	176	2	80 bytes
fxamacker/cbor 2.5	Encode (K)	576.5	272	4	80 bytes
ugorji/go 1.2.11	Encode (K)	1,408.0	1320	4	80 bytes
encoding/json	Encode (K)	850.9	368	4	129 bytes
encoding/json	Encode	670.5	336	4	179 bytes
encoding/gob	Encode	3,846.0	1,469	32	195 bytes

• Input: First example data in RFC 8392 Appendix A from a Go struct with keyasint tags.
• K = keyasint struct tag (or equivalent).   P = b.RunParallel().
• go1.19.7, linux_amd64, i7-12700H laptop. benchstat on go test -bench=. -count=20.

[x448]

This will show toarray and keyasint features shrinking encoded data size.

I can simulate the keyasint feature for encoding/json. Didn't see any struct tags for encoding/gob.

[fxamacker]

@x448 This is a great idea. However, I'm closing this (for now) so we can revisit and maybe reopen after v2.5.0 is released.

Comparisons in README can require time to update when any codec compared is updated.

Lack of time to update the README can be the only blocker to a new release being tagged, so we should reduce size of README first by removing or updating outdated comparisons that already exist.