feat: mutate ternary conditions

[espressolee]

Closes #196. Credit to @ryanfreckleton for #478 — that is where I started, but the tree has moved (node_mutation.py is now src/mutmut/mutation/mutators.py), so this is a fresh implementation rather than a rebase.

What was missing

A ternary is a branch nothing here mutated. Branch coverage does not see an uncovered arm either, so an untested arm read as a pass twice over. Before this, grep -rn "IfExp\|ternary\|if_exp" src/ tests/ docs/ returned nothing.

operator_if_exp yields two mutants per ternary, forcing it down each arm:

a if b else c  ->  a if (b) and False else c
                   a if (b) or True else c

The parentheses

The review on #478 asked to drop the precedence wrapping and use node.test directly. That is right for or True and wrong for and False, because and binds tighter than a top-level or. Over all 16 assignments of a if b or c else d:

mutant	differs from original
(b or c) and False	6 / 16
b or c and False	2 / 16, and only when b is falsy
(b or c) or True	2 / 16
b or c or True	2 / 16 — identical to the wrapped form

Unparenthesised, b or c and False parses as b or (c and False) and still takes the true branch whenever b is truthy, i.e. a mutant tests mostly cannot kill.

or True genuinely does not need the parentheses. I wrap both anyway so there is one rule instead of two, and it costs nothing — happy to make it asymmetric if you prefer.

Test impact

test_function_with_annotation shifts because its fixture contains a ternary: the two new mutants take numbers 1 and 2 and push the arithmetic and index mutants to 3–5. The three original assertions are kept and renumbered, and the two new mutants are asserted alongside them, so the test still shows nothing was lost.

Verification

pytest tests/ on 4f12080:

	failed	passed
before	12	345
after	12	345

Identical FAILED sets — I diffed them. The e2e inline-snapshot suite is unchanged.

tests/utils/test_safe_setproctitle.py is flaky on this machine regardless of the change (1 pass in 5 runs both with and without it), so it moves in and out of that list on its own; I checked rather than attributing it.

ruff check, ruff format --check and mypy all pass on the two changed files.