fix(vt): report cursor state a reset changes

[Rohilalala]

Fixes #936.

Screen.Reset assigns a zero Cursor rather than going through the setters that carry the callbacks, and by the time resetModes restores the mode table the values already match — so nothing reports the change:

emu.Write([]byte("\x1b[?25l")) // CursorVisibility(false)
emu.Write([]byte("\x1bc"))     // cursor is visible again, silently

Scope

The issue is about visibility. Position and style are the same bug in the same function — RIS sends the cursor home and clears its style just as silently — so this covers all three. Verified before the change: a cursor moved to (5,3) with a blinking-underline style reports 0 position changes and 0 style changes across a reset. Happy to trim it back to visibility alone if you would rather keep the change narrow.

I went with reporting rather than adding a getter, since it needs no new API and matches how the other cursor state is already observed. A CursorVisible() getter would also solve the issue if you would prefer that shape.

Why the callbacks are muted during the reset

Several paths inside a reset touch the cursor, and setAltScreenMode reports visibility unconditionally — even when it has not changed. Reporting each of them would turn one reset into a burst of changes, and a reset arriving from the alternate screen would report visibility twice. So the three cursor callbacks are muted for the duration and each real change reported once at the end.

Restoring them happens in a defer. Without it, a callback that panics mid-reset — AltScreen fires in there — and is recovered by the caller would leave the cursor callbacks nil permanently, silencing every later change. There is a test for exactly that.

Tests

• visibility reported once after a reset that changes it
• no report when the cursor was already visible
• exactly one report when the reset also leaves the alternate screen
• position and style each reported once
• callbacks still work after a panicking callback during a reset

Each was checked against a deliberately broken version to be sure it fails for the right reason: dropping the report, and turning the defer into a plain call at the end, each turn the matching test red.

go test ./... -race -count=3 -shuffle=on passes; golangci-lint run --config ../.golangci.yml ./... reports only the two findings already on main (emulator.go:421, csi.go:39), in files this PR does not touch.

Noticed, not fixed

setAltScreenMode reports CursorVisibility on every screen switch whether or not visibility changed, unlike setCursorHidden which checks first. That is what forced the muting here rather than a simple compare. It looked like its own change to make, but I am happy to fold it in.