add checkpoint_dir content-type, remove checkpoint variant

[bertsky]

This removes the older of the two parameterizations (a direct Calamari-like glob expression) in favour of the newer, more OCR-D-like model directory, and adds a proper format and content-type as required by the specs for optimal resmgr handling.

[codecov-commenter]

Codecov Report

Merging #70 (5fddd32) into master (76b34c5) will increase coverage by 0.97%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #70      +/-   ##
==========================================
+ Coverage   88.37%   89.34%   +0.97%     
==========================================
  Files           3        3              
  Lines         172      169       -3     
  Branches       39       38       -1     
==========================================
- Hits          152      151       -1     
+ Misses         11       10       -1     
+ Partials        9        8       -1     

Impacted Files	Coverage Δ
ocrd_calamari/recognize.py	88.95% <100.00%> (+1.00%)	⬆️

---

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 76b34c5...5fddd32. Read the comment docs.

[bertsky]

I had to fix the tests as well. We could also use kant_aufklaerung_1784-binarized instead of kant_aufklaerung_1784-page-region-line-word_glyph to get binary input without the need for ad-hoc IM binarization BTW.

[mikegerber]

It's a bit unfortunate to drop the former default parameter checkpoint, is there really no way to still support this and phase it out afterwards?

[bertsky]

It's a bit unfortunate to drop the former default parameter checkpoint, is there really no way to still support this and phase it out afterwards?

No, unfortunately I can't see any. We had to get to terms with our file vs directory resources and the semantics of content-type in a ocrd-tool.json – see related discussions in core. Unless we make the resource mechanism much more complicated (by specifying which params can have which resources), if we want to properly support directories as a resource type, individual processors should not have both.

[mikegerber]

I've also removed the checkpoint mention in the README in #80!