Closes #6540

In #6221 there was a mistake in calculating which folds need to be
re-ordered. When there are no folds after those that have been adjusted,
then `move_end` remains 0. This results in reverse_fold_order()
swapping folds that have been adjusted with uninitialised folds in the
remainder of the grow array.

Add a check in foldMoveRange() to account for this case.

Hope this will make people using feed_command less likely: this hides bugs.
Already found at least two:

1. msgpackparse() will show internal error: hash_add() in case of duplicate
   keys, though it will still work correctly. Currently silenced.
2. ttimeoutlen was spelled incorrectly, resulting in option not being set when
   expected. Test was still functioning somehow though. Currently fixed.

When foldUpdateIEMSRecurse() re-uses an existing fold, it misses the
case where the existing fold spans from before startlnum to after
firstlnum, the new fold does not span this range, and there is no
"forced start" of a fold. We add a case for this in.

Ensure that if there was no forced break in folds, we merge folds that
now touch each other.

Include testing for a tricky foldmethod=expr case that has never been a
bug. This case works at the moment because of some effects that are not
obvious when reading the code.
A test for this could be useful to ensure a regression doesn't happen.

vim-patch:8.0.0408