[abdull-ah-med]

• I have read CONTRIBUTING.md.
• I have created a discussion that was approved by a maintainer (for new features). (n/a, bug fix)

Fixes #333

What was wrong

With a width-10 textarea, typing 1234567890 keeps all ten characters on the first row, but typing 1 2 3 4 56 (also ten cells) moves 56 to the next row as soon as the 6 is typed. So a line of spaced words could only hold width-1 cells.

The cause is the final check in wrap(): >= width moved the last word to a new row whenever it made the line exactly full. A single unbroken word that fills the width takes a different path. It stays on its row, and only the trailing cursor cell goes to the next row.

What changed

• wrap() now keeps a last word that fits exactly on its row and puts only the trailing cursor space on the next row. This matches the unbroken-word case. The word still reflows once it grows past the width (1 2 3 4 567 → 1 2 3 4 / 567 ).
• I kept the reserved cursor row on purpose, so this does not overlap with Cursor gets stuck when navigating up/down in textarea #887 / fix(textarea): reach trailing wrapped cursor lines #1051.

Before / after (width 10, no prompt or line numbers)

input	before	after
1234567890	1234567890 / ``	unchanged
1 2 3 4 56	1 2 3 4 / 56	1 2 3 4 56 / ``
1 2 3 4 567	1 2 3 4 / 567	unchanged

How it was tested

• New TestWrapLastWordFillsWidth (table test of wrap, including a double-width rune) and TestTypingLastWordFillsWidthDoesNotReflow (typing through Update, then one more character to confirm the reflow). Both fail on main and pass with this change.
• LineInfo() row/column for the cursor at the end and at column 9 are now identical for 1234567890 and 1 2 3 4 56.
• go test ./..., go vet ./..., gofmt -l ., and golangci-lint run on the changed code all pass.

I prepared this change with AI assistance, then reviewed and tested it myself.