feat(ui): copy the comment at the cursor with Y

[Dhevenddra]

Fixes #569.

Problem

dd deletes the comment at the cursor, but there was no way to copy just that one. y exports the whole review, and visual-mode y copies diff text only: content_for_side (src/app/navigation.rs:540) returns None for comment annotations, so dragging over a comment box yanks nothing. Getting one comment out meant leaving the TUI for tuicr review comments --session ... | jq, which needs an id you cannot see from the box you are looking at.

Fix

Y in normal mode copies the content of the comment under the cursor.

• App::comment_content_at_cursor resolves through find_comment_at_cursor, the same lookup dd and i use, so Y and dd always agree on which comment the cursor is on. It carries over the side guard the delete path uses for line comments.
• The handler copies with copy_text_to_clipboard, so pbcopy, OSC 52, xclip and wl-copy behave exactly as they do for y, including the "(via terminal)" message.
• y and visual-mode y are untouched.

Remote threads are deliberately out of scope: they are read-only in v1 and their bodies live in forge_review_threads rather than the session. When the cursor is on one, the message names the forge instead of claiming there is no comment.

Help popup and docs/KEYBINDINGS.md both list the new key.

Evidence

cargo test:

test result: FAILED. 1300 passed; 2 failed; 2 ignored

The two failures are should_discover_worktree_with_relativeworktrees_extension and default_preference_routes_reftable_repo_to_cli. Both fail identically on an unmodified checkout of 8323f13 here, because this box runs git 2.43.0, which predates the relativeworktrees and reftable extensions. Nothing else fails.

New tests:

• should_yank_only_the_comment_under_the_cursor puts two comments on different lines and checks each one yanks its own content, not its neighbour's.
• should_yank_nothing_when_cursor_is_not_on_a_comment pins the None path on a plain diff line.
• should_map_lowercase_and_uppercase_y_to_separate_yank_actions pins y and Y to different actions.

cargo fmt --all --check and cargo clippy -- -D warnings are both clean. (clippy --all-targets reports a pre-existing items_after_test_module on src/handler.rs, unrelated to this change.)

One note on process: I could not dogfood this through the TUI itself, since I am on a Windows box driving the build through WSL2 without an interactive terminal for tuicr. The diff is small and self-reviewed, but I have not watched Y copy a comment in a live session, so a quick manual check on your side would be worth it before merge.

[agavra]

LGTM! Thanks @Dhevenddra