parse: Read text literal up to beginning of EndText ...

Using the end of the last word before EndText misses any trailing
whitespace, and it's a lost cause trying to figure out the offset
from after EndText given it might be whitespace or an end of life.

Instead use an offset from the start of EndText.

parse: Add code position to word value ...

This will be useful for finding the start of a word.

parse: Add Word class for lexer tokens ...

Instead of passing around a tuple, use a class.
Rename untokenized tokens to 'words' while we're at it.

parse: Fix tracing in lexer next() ...

This makes the line and column numbers accurate and prints
whitespace properly.

parse: Have lexer position be current instead of future ...

When running next(), self.pos and friends would actually be in reference
to the next call to next(), not the current one.

This is confusing, so change it to make sense.

parse: Fix text literals being cut off when placed at end of file

parse: Account for EOF when checking for whitespace ...

When parsing a file that doesn't end with whitespace (like a new line),
the parser would choke on having no symbols left. Fix that.

Fixes: 1519cef

parse: Decode file as UTF-8 ...

I was testing the parser with binary input.
Use Unicode instead and make sure we pass it that.

Add test harness and simple fuzz test

src: Reformat using black

Restructure files