Add options to allow overriding SendBufferSize pre process name prefix

[lingyuan2014]

This adds the ability to override SendBufferSize per process name prefix. I'm using prefix instead of regex because prefix seem to be sufficient for most usecases and is significantly faster:

➜  regexp_test go test -v -bench=.
goos: darwin
goarch: amd64
pkg: github.com/fiibbb/regexp_test
BenchmarkRegex
BenchmarkRegex-12                5526828               208 ns/op               0 B/op          0 allocs/op
BenchmarkPrefix
BenchmarkPrefix-12              344802882                3.46 ns/op            0 B/op          0 allocs/op
BenchmarkSuffix
BenchmarkSuffix-12              369626205                3.21 ns/op            0 B/op          0 allocs/op
BenchmarkContains
BenchmarkContains-12            142756165                8.41 ns/op            0 B/op          0 allocs/op
PASS
ok      github.com/fiibbb/regexp_test   7.896s

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Merging #772 into dev will increase coverage by 0.23%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##              dev     #772      +/-   ##
==========================================
+ Coverage   88.64%   88.87%   +0.23%     
==========================================
  Files          41       41              
  Lines        4140     4145       +5     
==========================================
+ Hits         3670     3684      +14     
+ Misses        356      349       -7     
+ Partials      114      112       -2     

Impacted Files	Coverage Δ
connection.go	88.73% <100.00%> (-0.35%)	⬇️
mex.go	80.56% <0.00%> (+0.94%)	⬆️
outbound.go	91.22% <0.00%> (+2.92%)	⬆️
root_peer_list.go	100.00% <0.00%> (+8.00%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 90ac437...4f86f67. Read the comment docs.

[prashantv]

Some minor comments, but looks good overall

[prashantv]

Looks good, when looking over the test cases, using a slice is definitely the right move, since the logic is deterministic. With a map[string]int, if you had both abc and abcd, we'd have non-deterministic results since map iteration order is undefined.

Left some small cleanup comments on the test, but otherwise LGTM.

[jaricftw]

LGTM, just nits to improve