Update to use Machine.py for USFM generation

[ddaspit]

There are a number of issues with USFM generation in the translate.py script:

• When translating on a USFM file with a table spanning multiple verses, the output USFM drops some verse markers and table formatting markers  #329
• ValueError: Some books failed to translate #444
• Orphan end markers are sometimes reported without a matching opening marker even though the opening marker exists. #379
• Translate script - parsing error on cross-references (\xt ... \xt*) in section reference lines (\sr) #367
• When translating a passage with embedded paragraph markers ("\pm"), verses are output in the reverse order #330
• When translating on a USFM file with a table spanning multiple verses, the output USFM drops some verse markers and table formatting markers  #329
• Reuse the same code for extracting verses where possible. #158

Most of these issues would probably be fixed if we scrapped the USFM parser that we currently use in the translate.py script and use the UsfmTextUpdater from Machine.py.

[isaac091]

Another known issue: \b markers immediately following chapter markers cause empty sentences to be added to the parsing output.

[isaac091]

A few more issues to check after switching parsers:

• Unclosed \fe breaks usfm parser #467
• Handle BOM (Byte-Order Mark) on custom stylesheets #338
• When there's nothing to translate, output nothing. #219 (also Translate script should not translate blank lines from the source text file #136)