coordinate arguments ought always accept -1 as "current cursor"

[dankamongmen]

See PR #2320 and issue #1696. We changed all length and coordinate arguments to unsigned for the listed functions. There are many other functions, however, which accept "-1" for any coordinate, meaning "current cursor position in the relevant dimension". This is a useful technique to have available, and we certainly must maintain it where already present.

Indeed, this ought be the global semantic for coordinates. So on the affected functions, change those coordinate params back to int from unsigned, and implement the "-1" case everywhere. Lengths remain unsigned; that's fine. This change never got released, so it's not a big deal user-wise (and it ought hew more closely to the current ABI, which i was admittedly a bit sloppy about in #2320).

Don't cut 2.4.9 until this is done.