Perf improvement

[nitely]

Changes:

• Preallocs definite strings (cbor string/byte-string) and seq (cbor array)
• Special care taken to validate length before alloc

TODO:

• Sensible default limits: so a cbor item with a length of int64.high won't cause an OOM
• See if can overload decode to set length limits to input len when config not passed. Requires nim-serialization >= 0.5.2 Fix decode params forwarding status-im/nim-serialization#113
• Overflow check for uint64 value sum passed to setLen
• Tests for the attack vector
• Perf improvement in orc status-im/nim-faststreams#86 companion PR (bottleneck)
• Do "write" sanity checks on debug mode and some cborSerializationWriteChecks define; follow up to Write text/byte-string in chunks and byte-by-byte #6

this is ~3x faster on Orc and ~2x faster on Refc

[github-actions]

[nph] _{reported by reviewdog 🐶}

Suggested change

	template parseArrayLikeImpl(p: var CborParser, majorExpected: CborMajor, body: untyped) =
	template parseArrayLikeImpl(
	p: var CborParser, majorExpected: CborMajor, body: untyped
	) =

[github-actions]

[nph] _{reported by reviewdog 🐶}

Suggested change

	template streamVar: untyped = w.stream
	template streamVar(): untyped =
	w.stream