Allow including compression level when training a dictionary

[mortengrouleff]

The compression improves when the level for the training is close to the level for the compression step.
(100% Compatible for the public API.)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 59.70%. Comparing base (c76455c) to head (c51c9c3).
Report is 24 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #326      +/-   ##
============================================
- Coverage     60.01%   59.70%   -0.31%     
- Complexity      308      314       +6     
============================================
  Files            26       27       +1     
  Lines          1473     1504      +31     
  Branches        170      175       +5     
============================================
+ Hits            884      898      +14     
- Misses          434      448      +14     
- Partials        155      158       +3     

[mortengrouleff]

@luben Would you consider this PR for merge and potential patch-version bump?

[mortengrouleff]

Bump?

[mortengrouleff]

Why use the desired compression level?

https://github.com/facebook/zstd/blob/63779c798237346c2b245c546c40b72a5a5913fe/lib/zdict.h#L237

[luben]

Thanks for the contribution! I was away for 2 weeks - late summer vacation - and didn't check github.

[luben]

released with v1.5.6-6

[mortengrouleff]

Thanks for the contribution! I was away for 2 weeks - late summer vacation - and didn't check github.

Thanks!