style: apply Ruff's remaining automatic fixes

[mkb79]

Picks up the 110 findings #282 left behind because they need --unsafe-fixes. Ruff calls them unsafe where a fix can alter meaning rather than only layout, so each was read rather than trusted.

	before	after
findings	230	121
file/rule pairs pinned in the baseline	75	36
files with findings	23	17

⚠️ audible --help changes

Two thirds of the fixes are D415, a full stop at the end of a docstring's first line. Fifteen of those docstrings belong to click commands, which use them as help text:

 activation-bytes  Get activation bytes.
-api               Send requests to an Audible API endpoint
+api               Send requests to an Audible API endpoint.
-library           Interact with library
+library           Interact with library.

activation-bytes already ended in a full stop, so this makes the list consistent rather than changing its style. Compared before and after on the real CLI; the added stop is the only difference.

Whitespace inside strings

W291/W293 strip trailing whitespace, which Ruff marks unsafe because whitespace inside a string is content. 24 of these lines are exactly that — the quickstart intro, two click help texts and several docstrings.

None of it shows. The intro goes to secho() unchanged and a space before a line break is invisible in a terminal; click rewraps help text regardless. audible api --help differs only by the added full stop, confirmed by diffing the output.

Two that are neither cosmetic nor a rewrite

• _logging.py gains stacklevel=2 on a warn() call, so the warning points at the caller instead of at the logging module. Small, but a real change in what the user sees — and the reason the rule exists.
• plugins.py has its %-formatted plugin warning turned into an f-string. Same text.

Nine ordinary rewrites

A set comprehension instead of set() around a generator, a bare function reference instead of a lambda that only forwarded to it, a list comprehension instead of map() with a lambda, [*initial, current] instead of initial + [current], list(range(...)) instead of a comprehension over it, and {} instead of dict().

is_supported_file is a classmethod, so dropping the lambda keeps the same behaviour, and nothing in the code base shadows dict.

What is left

121 findings that need judgement rather than a tool: 78 G004 (f-strings in logging calls, which are evaluated even when the level discards them), 12 PLR0917, 10 D205, 6 S101 and a handful of others. G004 is the obvious next piece and deserves its own pass.

Verified

ruff check . and ruff format --check . pass, 56 tests green, and against the live API library list --resolve-podcasts --start-date 2008-01-01 returns the identical result set on this branch and on master in the same window — 918 lines, sorted diff empty.