chore: clear the clippy warnings that were real (#32)

[jspaezp]

Twelve of the nineteen warnings; the other seven need a decision rather than an
edit.

Rebased onto main now that #98 has merged, so this is a single commit again.
One semantic fixup folded in: #98 added a test asserting on
DiannReadingError::DiannPrecursorParsingError, which this PR renames to
PrecursorParsing. Git rebased cleanly and the breakage only showed up at
compile time.

Fixed

• very_complex_type x6 -- the three timscentroid benchmark examples spelled
  ((f32,f32),(f32,f32),(f16,f16)) six times between them. Each gets a BenchQuery
  alias.
• enum_variant_names x3 -- Diann/Skyline/SpectronautReadingError suffixed
  every variant with Error, so a match arm read DiannReadingError::IoError.
  Suffix dropped.
• should_implement_trait -- DecoyMode::from_str was an inherent method
  shadowing the trait, so "reverse".parse::<DecoyMode>() did not work at all. It
  implements FromStr now, and the one call site uses .parse().
• mut_range_bound -- speclib_build_cli's digest loop hand-rolled a counter
  next to the iterator; it uses enumerate.

The one to look at

PossibleAggregator's variants carried the same redundant suffix, but it derives
clap's ValueEnum, so the variant names are the CLI values. A plain rename
would have silently changed --aggregator from chromatogram-aggregator to
chromatogram.

Renamed with the strings pinned by #[value(name = "...")], so the surface is
identical and now stated rather than implied. --help still offers exactly
point-intensity-aggregator, chromatogram-aggregator, spectrum-aggregator,
and the Carafe end-to-end test passes chromatogram-aggregator, so a future
rename that forgets the pin fails a test instead of breaking a caller.

Not fixed: 7 x too_many_arguments

Functions of 8-10 parameters. Silencing them with #[allow] would be suppression,
and bundling all seven into config structs is a real refactor with real regression
risk -- so it wants a call rather than my guess. They split two ways:

function
PyElutionGroup::new (pyo3 #[new])	the parameter list is the Python signature
render_chromatogram_plot (egui)	flat render args are idiomatic there
run_pipeline, execute_pipeline, process_single_file	would genuinely read better as a config struct
process_and_serialize, build_entry	same

Happy to do the config-struct refactor for the bottom five as a follow-up.