Degenerate cases with initial spellcheck

[caffodian]

Following cases are very bad:

• One enormous block
• Multiple blocks with identical content (there is a bug, somewhere, and the words get re-checked)
• many tiny blocks (it should batch by number of words, not number of blocks, but we were in a hurry at the time and that fixed most real world cases.)

[caffodian]

It makes sense to break out the enormous block separately.

The others can be solved pretty easily, but because the smallest common unit you can use for scanning/marking is a block, it would take a fair bit of refactoring to fix the enormous block case.

[caffodian]

Fixing the many tiny blocks should be a priority due to tables and lists.