perf(encoding): speed up EncodeVarint with io.ByteWriter+hand rolled varintEncode (backport #917)

[mergify]

This change speeds up EncodeVarint by testing if the input writer implements io.ByteWriter and if so, goes to use our hand-rolled varint encoder, instead of using the awkward standard libary encoding/binary.PutVarint that requires a byteslice, which we also retrofitted using a bytearray pool.
While here, added parity tests to ensure that we get the exact same results as with the Go standard library's encoding/binary package with caution from https://cyber.orijtech.com/advisory/varint-decode-limitless and also added benchmarks whose results reflect the change in just the benchmark initially

$ benchstat before.txt after.txt
name            old time/op    new time/op    delta
EncodeVarint-8     360ns ± 3%     245ns ± 3%  -31.80%  (p=0.000 n=10+10)

name            old alloc/op   new alloc/op   delta
EncodeVarint-8     0.00B          0.00B          ~     (all equal)

name            old allocs/op  new allocs/op  delta
EncodeVarint-8      0.00           0.00          ~     (all equal)

Fixes #891

---

This is an automatic backport of pull request #917 done by Mergify.