perf(mangling): optimized the processing of initialisms

[fredbi]

This PR significantly improves the performance of name-mangling
utilities (e.g. ToGoName, etc).

The need for this occurred while benchmarking go-swagger's CI suite:
surprisingly, the topmost allocator was swag.ToGoName.

It is the result of a dozen successive optimization passes driven
by profiling.

These functions now execute ~10x faster and need 100 times less
memory allocations. See BENCHMARK.md

Optimization techniques used to reduce allocations:

• pointer -> value (everything was pointers: now everything is values)
• interface -> struct with a Kind field
• closure -> func
• string to []byte: use unsafe conversion
• string concatenation -> use bytes.Buffer (recyclable, unlike
  strings.Builder)
• static values converted over and over again: pre-bake the conversions
• var x []T -> make([]T, 0, heuristic size)
• temporarily allocated values -> use pools to recycle previously
  allocated data items

Optimization techniques used to reduce CPU:

• read Unicode rune -> short-circuit for single byte runes
• map lookup -> func with switch statement
• for i, v := range -> for i := range (minor impact)

Signed-off-by: Frédéric BIDON fredbi@yahoo.com

[codecov]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (b3e7a53) 85.45% compared to head (0ba0453) 86.91%.

❗ Current head 0ba0453 differs from pull request most recent head 5904b43. Consider uploading reports for the commit 5904b43 to get more accurate results

Files	Patch %	Lines
util.go	89.85%	6 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #79      +/-   ##
==========================================
+ Coverage   85.45%   86.91%   +1.45%     
==========================================
  Files          12       13       +1     
  Lines        1609     1842     +233     
==========================================
+ Hits         1375     1601     +226     
- Misses        202      208       +6     
- Partials       32       33       +1     

Flag	Coverage Δ
oldstable	86.91% <98.33%> (+1.45%)	⬆️
stable	86.91% <98.33%> (+1.45%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fredbi]

integration test w/ go-swagger: go-swagger/go-swagger#3060

[fredbi]

I've been testing this over the week with go-openapi/analysis, go-openapi/validate and go-swagger.
The reduction in generated garbage is quite significant, but not enough to translate in noticeable timings improvements.
Consider this PR as a supplement to the effort started with go-swagger/go-swagger#3063 and another important forthcoming PR on go-openapi/validate, as joint efforts to contribute to go-swagger/go-swagger#2649