use pie-core instead of pytorch-ie (core only, not for datasets)

[ArneBinder]

This PR also:

• fixes use EnterDatasetMixin, ExitDatasetMixin, EnterDatasetDictMixin, and ExitDatasetDictMixin from pytorch-ie (or pie-core) #174
• decreases the patch codecov target to 85%

requires:

• remove dependency on TransformerSpanClassificationTaskModule from pytorch-ie #180
• make document casting relaxed regarding annotation types #181
• use pie_modules.utils.sequence_tagging #179
• release of pie-modules including use pie-core implicitly pie-modules#176:
  • so that the document and annotation types imported via pie_modules are derived from pie_core.Document
  • this is required for tests

TODO:

• use use pie-core implicitly pie-modules#176
• add annotations required for common base dataset builders, i.e, replace BinaryRelation, LabeledMultiSpan, LabeledSpan and TextBasedDocument with Brat variants
• fix tokenization issues EDIT: was caused by pie-modules version below fix token_based_document_to_text_based / text_based_document_to_token_based for dependent predicted annotations  pie-modules#135
• fix "number of annotations mismatch" in sciarg
  • was caused in resolve_parts_of_same variant because SpansViaRelationMerger requires spans of type LabeledSpan, but we now use BratSpans in the BratDocuments
  • fixed in 221c62b by converting documents before applying SpansViaRelationMerger
• adjust readme

Follow-ups:

• use pie-core in dataset builders #184
• remove torch from dev dependencies (requires remove pytorch-lightning dependency pie-core#52)

context: ArneBinder/pie-core#17

[codecov]

Codecov Report

Attention: Patch coverage is 87.05882% with 11 lines in your changes missing coverage. Please review.

Please upload report for BASE (main@61575b4). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/pie_datasets/builders/brat.py	85.71%	11 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #178   +/-   ##
=======================================
  Coverage        ?   92.52%           
=======================================
  Files           ?       10           
  Lines           ?      963           
  Branches        ?        0           
=======================================
  Hits            ?      891           
  Misses          ?       72           
  Partials        ?        0           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.