Feature: Autodetect valid word lengths

Autodetect valid word lengths

With Heuristic exclude suggestions, I've started reviewing what's worth checking.

For one, a significant portion of the two letter space is already in the dictionary.

For another, programmers often develop two letter sequences, either as abbreviations or identifiers.

In a recent pull request of mine, every single 2 letter pair that was flagged was valid.

I know of one repository with a run of as that exceeds 32766 characters.

The checker (while reading the dictionary) will find the shortest and longest word lengths.

• Unless overridden, it will ignore things that are shorter than the shortest word length (at least 3).
• Unless overridden, it will ignore things that are longer than two characters longer than the longest word (this allows one to accidentally add two characters to a long word and still be flagged for a typo).

There are now two flags to configure the length of "words":

• shortest_word -- you can set this to "2" to restore the old behavior
• longest_word -- you could set this to some obscene number like "32767" to get closer to the old behavior