TLDR-599 re-label datasets (#412)

* TLDR-616 add parsing parameters saving to tasker, fix bugs

* TXT images creator fixed

* Fix tasker for diplomas

* Fix tests, speed up txt_images_creator

.github/workflows/test_labeling.yaml

@@ -1,4 +1,4 @@
-name: CI
+name: CI labeling
 
 # Controls when the action will run. 
 on:

dedoc/structure_extractors/abstract_structure_extractor.py

@@ -54,6 +54,9 @@ def _postprocess(self, lines: List[LineWithMeta], paragraph_type: List[str], reg
         :param excluding_regexps: list of filtering garbage regular pattern according to list of paragraph types
         :return: new post-processed list of LineWithMeta
         """
+        if self.config.get("labeling_mode", False):
+            return lines
+
         result = []
         for line in lines:
             if line.metadata.hierarchy_level.is_raw_text() and len(line.line) == 0:  # skip empty raw text