Don't redraw a blank line just because its style differs

[afonsojanu]

Fixes #2008

get_max_column_index already ignores trailing unstyled whitespace when figuring out how far along a line has real content, but that only decides how far the column loop goes. It doesn't stop a fully blank line's single remaining cell from being redrawn anyway, since a freshly reset previous screen's default cell has a different (transparent) style than whatever the new blank cell got assigned, so the diff check treats it as changed even though nothing visible actually is. That's the stray space the issue describes, e.g. the bottom-left corner right after starting something like ipython in an empty terminal.

The one-line fix suggested in the issue (skip drawing when a line's max column index is 0) doesn't quite work: a line whose only visible character sits at column 0 also has a max index of 0, since get_max_column_index returns the highest index with real content and there's only one column to check. Applying that fix as written would silently drop any single-character line. This instead tracks whether a line has any real content at all, computed once per row the same way get_max_column_index already does, and only skips the redraw when there's genuinely nothing to show.

Added tests/test_renderer.py (there wasn't one before) exercising _output_screen_diff directly against a capturing output, covering: a blank line no longer gets a space written, a line whose only character is at column 0 still gets drawn, and a line with content after column 0 is unaffected either way. Verified the first case fails without the fix (a bare space shows up in the output) and the second would have failed against the naive one-line version.

Ran the full test suite (159 passed) plus ruff format --check and ruff check on the touched files.