Interesting question. The infrastructure is all there, but the code to do that actual task doesn't exist. The approach I would take is:

1) Parse the FLV with existing code.
2) grep for keyframes at about the time when you want to cut
3) Write new code to copy the frames to one FLV or another, depending which side of the time cut they're on. You'll have to create a new holder for each.
4) Write new code to generate a header for each FLV (cribbing heavily from the existing header writing code)

All in all, a challenging but not a super-time-consuming task -- maybe a day's work if you have some familiarity with the FLV file format, or 2-3 days if you're starting from zero.

Or, contact me offline if you can fund the addition of this feature to the library. :-)