Filename => codepoint conversion is over-lenient

[rsheeter]

Discovered courtesy of googlefonts/tofu#1, if you ask nanoemoji to compile a file names tofu.svg it runs to completion. It should fail; this matches neither of our naming schemes.

nanoemoji/src/nanoemoji/codepoints.py

Lines 22 to 26 in fb4b0b3

	def from_filename(filename):
	match = regex.search(r"(?:^emoji_u)?(?:[-_]?([0-9a-fA-F]{1,}))+", filename)
	if not match:
	raise ValueError(f"Bad filename {filename}; unable to extract codepoints")
	return tuple(int(s, 16) for s in match.captures(1))

is using search when it should be ensuring it matches the entire filename without extension (e.g. https://docs.python.org/3/library/pathlib.html#pathlib.PurePath.stem)