perf: optimize factory performance

[muktihari]

• Use [256]proto.Field in the predefined message in the factory for fast look up, this ensure O(1) look up when creating Field instead of looping to get the Field like previous version.
• Since the predefined messages in factory is now using fixed array, the behavior of CreateMesg is slightly changed since we need to populate the slice for the returned message's fields. It will populate the data once during the first invocation, since this method is not used either by the decoder or the encoder it should be fine, method documentation is updated to make users aware about this change in case users want to use this method.

goos: darwin
goarch: amd64
pkg: github.com/muktihari/fit/decoder
cpu: Intel(R) Core(TM) i5-5257U CPU @ 2.70GHz
         │   old.txt   │               new.txt               │
         │   sec/op    │   sec/op     vs base                │
Decode-4   126.4m ± 1%   108.6m ± 1%  -14.08% (p=0.000 n=10)

         │   old.txt    │            new.txt             │
         │     B/op     │     B/op      vs base          │
Decode-4   67.39Mi ± 0%   67.39Mi ± 0%  ~ (p=0.361 n=10)

         │   old.txt   │             new.txt             │
         │  allocs/op  │  allocs/op   vs base            │
Decode-4   900.0k ± 0%   900.0k ± 0%  ~ (p=1.000 n=10) ¹
¹ all samples are equal

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.02%. Comparing base (b531a82) to head (ece2daa).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #120   +/-   ##
=======================================
  Coverage   99.02%   99.02%           
=======================================
  Files          45       45           
  Lines        4396     4396           
=======================================
  Hits         4353     4353           
  Misses         43       43           

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