`Key` has `Delete`, `PageUp` and `PageDown` but no `Insert`

[vyncint]

Good first issue. Three short match arms and two test lines, with every
neighbour already written for you to copy.

Today

Key (crates/termlens/src/keys.rs:15) covers the navigation cluster — almost:

Delete    ESC [ 3 ~     keys.rs:100
Home      ESC [ H       keys.rs:105
End       ESC [ F       keys.rs:106
PageUp    ESC [ 5 ~     keys.rs:107
PageDown  ESC [ 6 ~     keys.rs:108

Insert — ESC [ 2 ~ — is missing. It is the 2 in a run that already has
3, 5 and 6.

A test for an application that binds Insert (a table editor toggling insert
mode, a file manager's "select and move down", vim-likes) cannot send it. The
closest workaround is Key::Char with a hand-written escape, which is not
something a test should have to know.

Fix

Three places, each with a Delete or PageUp line immediately next to it:

1. The variant, keys.rs:48 area — beside Delete, with a doc comment in
   the same style: /// Insert (ESC [ 2 ~).
2. Key::encode, keys.rs:100 area:

   Key::Insert => b"\x1b[2~".to_vec(),

3. chord_base, keys.rs:311 area — this is what makes
   Key::Insert.shift() and friends encode correctly:

   Key::Insert => ChordBase::Tilde(2),

Not encode_modal (keys.rs:161). That handles DECCKM, which only applies
to the cursor keys and Home/End; Insert is unaffected and falls through to
encode correctly on its own. Worth a moment to convince yourself of that rather
than taking my word for it.

Tests

Two existing tables in keys.rs, one line each:

• xterm_encodings (keys.rs:360): (Key::Insert, b"\x1b[2~")
• chord_encodings (keys.rs:408): something like
  (Key::Insert.shift(), b"\x1b[2;2~") — check the modifier arithmetic against
  the Delete.ctrl() line right above it rather than trusting mine.

There is also an end-to-end path if you want it. tests/fixtures.rs:62 sends
keys to the form-echo fixture and asserts on the name it echoes back. The
fixture's describe (fixtures/form-echo/src/main.rs:90) has an arm per key and
currently has none for KeyCode::Insert; adding one and a case in the test table
proves the byte sequence survives a real PTY and a real input parser, not just
our own table. Optional, and a good stretch if the first part goes quickly.

One thing to know before you start

Key is not #[non_exhaustive] today, so adding a variant is a breaking
change — cargo-semver-checks runs at release and will say so. That is why
this sits in the v0.7 milestone rather than a patch.

#165 proposes making Key #[non_exhaustive], which would take that cost off
every future key. The two issues are independent — take either, in either order —
but if you fancy doing both, doing #165 first makes this one free from then on.

Done when

Key::Insert encodes ESC [ 2 ~, chords with modifiers correctly, is covered by
the encoding tables, and has a line under [Unreleased] in CHANGELOG.md.

Getting started

CONTRIBUTING.md has the setup and the DCO sign-off (git commit -s); AI help is
welcome, AI attribution is not. crates/termlens/src/keys.rs is 483 lines and
reads top to bottom — you can hold all of it at once, which is why this is a good
place to start. Ask here if anything is unclear.