fix: use the standard 2.54mm JEDEC pitch for to220 and honor the p parameter

[Furox-Art]

/claim #790

Problem

to220 derived its lead pitch from the plastic body width:

const computedPitch = Math.max(2.5, (w * 0.4) / (numPins - 1))

• At the default w=13mm, 3 pins → 2.6mm, 0.06mm off per step from the fixed JEDEC TO-220 pitch of 0.1in (2.54mm)
• The pitch shifted whenever body width changed, but a TO-220 lead pitch is fixed regardless of body size
• The declared p parameter (default "5.0mm") was parsed and never read — to220_p2.54mm had no effect

Fix

Mirror the sibling to220f, which already hardcodes 2.54mm with the comment "to match KiCad":

• p defaults to 2.54mm and is the single source of the pitch (const pitch = mm(parameters.p))
• Body width no longer influences hole positions
• parameters.p reports the effective pitch

Tests (6 new, snapshot-free, deterministic)

Check	main	this PR
to220_3 default pitch	2.6mm	2.54mm
p: "5mm" override	ignored	honored → gaps [5, 5]
w: "25.4mm" body	pitch scales	still 2.54mm
to220_4 centered	✓	✓ (unchanged)
to220_5 pitch	derived	2.54mm
reported parameters.p	computed	effective 2.54

Snapshots

The 4 affected to220_*.snap.svg snapshots were regenerated on ubuntu-latest CI (via a temporary workflow_dispatch helper, removed before this PR) — no local-font drift can enter the pixel comparison. Every other snapshot in the repo is untouched (verified: diff touches exactly the 4 to220 SVGs).

Note: #828 also targets this issue but bundles fixes for three unrelated issues (#786/#788 fpc/pad defaults, solderjumper) into one PR; this PR is scoped to #790 only and includes the pitch-regression tests above.

[Furox-Art]

Fresh local run of the regression suite (2026-09-20, bun 1.4.2, this branch):

bun test tests/to220-pitch.test.ts

 6 pass
 0 fail
 12 expect() calls
Ran 6 tests across 1 file. [1.26s]

Covers the 2.54mm TO-220 pitch geometry: pin pitch, pad size, body offsets and the regression snapshot for to220("TO220_2").