fix(keybindings): allow "." to be bound from keybindings.toml

[nocstah]

The period key cannot be named anywhere in the keybinding system: it is
absent from KebindingResolver._namedKeys, _canonicalKeyName() rejects it (it
only special-cases "/", ":", "?" and ""), and _eventKeyName() returns null
for Qt.Key_Period.

The result is a silent no-op rather than an error. A user writing

[keybindings]
toggle_hidden = "."

gets an entry that parses to null, is dropped as invalid, and never binds --
indistinguishable from a typo, and with nothing in the log to say why. Since
P2.5's whole premise is that alternative layouts can move any key they like,
a punctuation key that cannot be expressed at all is a real gap.

Fixes it the same way the four existing punctuation keys are handled:
_namedKeys gains "period"/"." -> Qt.Key_Period, _canonicalKeyName() accepts
".", and eventKeyName() prefers the symbol form -- that last part matters,
because both spellings map to one Qt.Key*, and only one canonical string may
come out of _eventKeyName() or it can never compare equal to what
_canonicalKeyName() produced.

Adds a regression check ("a punctuation key (".") can be bound from config")
covering the real user path: the override is accepted, "." reaches
toggle_hidden, its previous Ctrl+H stops working per the full-replacement
policy, and "/" -- which always worked -- still searches. Verified to fail
without the fix: override accepted=false "."->toggle_hidden=false.

CheckKeybindings' own NAMED_KEYS also needed "." so the harness can express
it: it is deliberately independent of the resolver's table, and its
single-character fallback computes Qt.Key_A + (charCode - 'a'), which for "."
is Qt.Key_A - 51.

Selfcheck on this branch: 125 passed, 1 failed, 126 total. The one failure is
the pre-existing thumbnail-video.sh check, which fails the same way on an
unmodified master build here (it needs ffmpegthumbnailer, which this machine
does not have -- see the separate PR).

Same class of gap remains for other unlisted punctuation (",", ";", "-",
"="...). Happy to generalize _canonicalKeyName() to accept any single
printable ASCII character instead of an allow-list if you would prefer that
shape.