fix(to220): standard 2.54mm lead pitch and KiCad TO-220-3 pads

[zkasuran]

Summary

to220 now emits the KiCad TO-220-3_Vertical land pattern: a fixed 2.54mm lead pitch, a rectangular pin-1 pad and pill lead pads. Previously it placed plain circular holes on a body-width-derived pitch.

Fixes #790.

The bug

The pitch was computed from the body width:

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

For the default 3-pin part (w = 13mm) that gives 2.6mm, not the JEDEC TO-220 pitch of 0.1 inch (2.54mm). It also drifts with the body width and left the declared p parameter unused. The to220f sibling already hardcodes 2.54mm with the comment TO-220F uses 2.54mm standard pitch to match KiCad, so to220 was the odd one out.

The fix

Use a fixed 2.54mm pitch (now honoring p) and build the KiCad TO-220-3 land with platedHoleWithRectPad for pin 1 and platedHolePill for the rest, the same helpers to220f uses. to220f itself is untouched: it still overrides the holes for its own 1.2mm drill.

Parity (repo harness vs Package_TO_SOT_THT.pretty/TO-220-3_Vertical)

	before	after
avgRelDiff (copper)	36.52%	0.00%
courtyardDiffPercent	17.02%	11.77%

The copper now matches KiCad exactly. The residual courtyard delta is the shared TO-220 body outline: KiCad draws a slightly tighter courtyard for the metal-tab TO-220-3 than for TO-220F, while footprinter uses one body courtyard for both variants (to220f still matches TO-220F-3 within 0.5%).

Tests

• new tests/kicad-parity/to220_3_kicad_parity.test.ts asserting avgRelDiff < 0.01 plus courtyardDiffPercent < 13, with SVG and boolean-difference snapshots
• refreshed the to220_2/3/4/5 render snapshots

Verified locally

• bun test: 522 pass, 0 fail
• bun run build: success
• bunx biome format .: clean

AI disclosure

AI assistance (Claude, Anthropic) was used in developing this change. The design, review and verification were done by the author. Verified locally before submitting: bun test (522 pass, 0 fail), bun run build, bunx biome format ..

[techmannih]

Please add KiCad tests for the footprints you changed.

[zkasuran]

Added tests/kicad-parity/to220_4_kicad_parity.test.ts in db11e19, matching the to220_3 parity that was already in the PR. Against Package_TO_SOT_THT.pretty/TO-220-4_Vertical the copper matches KiCad exactly (avgRelDiff 0.00%) with courtyardDiffPercent 11.77%, the same residual body-outline delta as to220_3, so it reuses the same < 13 courtyard threshold and ships the SVG plus boolean-difference snapshots.

That gives KiCad parity for both fixed 2.54mm-grid variants (to220_3 and to220_4). I deliberately did not add parity for the other two changed footprints, because their KiCad references are not on a 2.54mm pitch and a passing assertion there would need a threshold that certifies nothing:

• TO-220-2_Vertical is a 5.08mm land (the two outer leads of the frame), so the fixed 2.54mm part measures avgRelDiff 29.40% / courtyardDiffPercent 17.15% against it.
• TO-220-5_Vertical is the 1.7mm Multiwatt land, so 2.54mm measures avgRelDiff 36.03% / courtyardDiffPercent 24.30%.

Both keep their render snapshots in tests/to220.test.ts. If you want real KiCad parity across every pin count I can follow up with a per-pin-count pitch (5.08mm for the 2-pin, 1.7mm for the 5-pin) so to220_2 and to220_5 land on their KiCad packages too. That is wider than #790, which is why I scoped this PR to the 2.54mm land. Happy to fold it in here or open a separate PR, your call.

bun test on the to220 files is green (9 pass) and bun run build succeeds.