feat: decoder preallocate messages

[muktihari]

Add decode option to precalculate messages first before the actual decoding to avoid re-allocation when appending each decoded message. This is only eligible if the given io.Reader is an io.ReadSeeker and WithBroadcastOnly() is unset.

How fast it is may depends on I/O performance as we need to read the file twice, but typically, it beats the cost of re-allocation especially when the FIT file contains so many messages, e.g. more than 32k messages (I measure up to 237k messages from an ultra-cycling and ultra-run activities), the more the messages we have, the greater the benefit we'll gain. But more measurements are needed since users' cases may vary. I decide to not enable this by default since I'm not confident enough that this will cover most cases and probably leaving this as an opt-in feature is the best decision, we'll see.

This PR is open for feedback. Please feel free to share your thoughts.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 8.69565% with 84 lines in your changes missing coverage. Please review.

Project coverage is 98.21%. Comparing base (078b1cc) to head (b5ab6d7).

Files with missing lines	Patch %	Lines
decoder/decoder.go	6.74%	82 Missing and 1 partial ⚠️
decoder/readbuffer.go	66.66%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##            master     #493      +/-   ##
===========================================
- Coverage   100.00%   98.21%   -1.79%     
===========================================
  Files           42       42              
  Lines         4618     4708      +90     
===========================================
+ Hits          4618     4624       +6     
- Misses           0       83      +83     
- Partials         0        1       +1     

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

[muktihari]

Test coverage will be added later once everything is settled.

[muktihari]

I will drop this since the benefits are considered small compared to the complexity it adds to our code. Additionally, having too many options will create a headache for users who are just starting to learn how to use this library. So let's keep it simple for now.

With our existing building blocks (Decoder, MesgListener, and RawDecoder), users have the ability to implement this approach on their own. However, they can also request this feature by providing a clear rationale for why it should be enabled.