fix: handle OSC sequences split across buffer chunks (#11144)

The OSCProcessor assumed OSC sequences (like OSC 52 clipboard) arrive in a single
buffer chunk. However, when sequences exceed 1024 bytes, they get split across
multiple chunks, causing the prefix and suffix to be in different calls.

This fix adds buffering to accumulate incomplete OSC sequences until the terminator
(BEL or ST) arrives, allowing proper handling of arbitrarily long OSC 52 clipboard
data and other long OSC sequences.

Fixes #6001

Co-authored-by: ToastyTheBot <toastythebot@anthropic.com>