PDFs with large spacing between letters are not properly parsed

[alephdata]

PDFs with large spacing between letters are not properly parsed

[stchris]

Testing the fix in #402 I realized that there is a compromise to be made, see #403. The issue is with small text versus spaced apart text. I consider the former more important, so this is how it looks:

With the previous word_margin of 0.27 small text like this:

would be indexed like this:

and the food menu with the large spacing:

would be indexed quite decently:

But with word_margin=0.2 small text works fine:

and the food menu is ... still quite decent?

[stchris]

I prefer the above changes from #403 (word_margin=0.2) and will roll with those. I'm happy about having extended the test suite in terms of small text.