Generate multiple lstmf files with a single tesseract invocation

[lorenzob]

Environment

• Tesseract Version: 4.0.0
• Platform: Linux acqua 5.3.0-40-generic Fix potential null pointer dereference in ccmain/paragraphs.cpp. #32~18.04.1-Ubuntu SMP Mon Feb 3 14:05:59 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux

Current Behavior:

The tesseract command line tool is able to generate only one lstmf file at a time.

When using single line images for training this means to generate 20/50k lstmf files or more and this can take hours.

Expected Behavior:

It should be possible to generate lstmf files for a at least a few thousands files without incurring in the overhead of starting a new tesseract process for each file.

I would expect the execution time to be reduced to a few minutes if not seconds.

Suggested Fix:

The executable should accept a list of files (or a folder and process all the tif files found there).

[stweil]

I wonder whether lstmf files are useful at all. The training process could create the necessary data in memory on the fly instead of reading lstmf files.