[NFC] Trim MIFARE Plus generator & admin-key mapping (in-place cleanup)

[mishamyte]

What

Two behavior-preserving, in-place cleanups of the MIFARE Plus code. No functional change, no public SDK API change (targets/f7/api_symbols.csv untouched).

1. Data-drive the MIFARE Plus manual-add generators (lib/nfc/helpers/nfc_data_generator.c)

   • The 18 near-identical MIFARE Plus variants were 18 one-line thunk functions + 18 handler-table entries. They differ only by parameters (UID length, product type, memory size, ATS historical bytes), so they now come from one data-driven mf_plus_generator_configs[] table, dispatched via a small range predicate. The Ultralight/NTAG and Classic types keep their bespoke handlers.
   • Dropped mf_plus_ats_hist_ev, which was byte-identical to mf_plus_ats_hist_s (EV1/EV2 are identified by GetVersion, not the ATS).
   • A _Static_assert pins the config table to the enum range so a future type can't desync it.

2. Single source of truth for the admin-key addresses (lib/nfc/protocols/mf_plus/*)

   • The admin-key address↔type mapping (0x9000 CardMaster, 0x9001 CardConfig, 0x9003 L3Switch, 0x9004 SL1CardAuth) was copied in 3 places — the poller's mf_plus_admin_key_id[] array and two identical switch statements in the listener. Folded into one mf_plus_admin_key_addresses[] table with a forward helper (mf_plus_get_admin_key_address) and a reverse helper (mf_plus_admin_key_type_from_address). Internal helpers only.

Why

Follow-up to #1033. The original plan there was to relocate the whole generator out of firmware into the NFC app to reclaim multiple KiB of internal flash. That turned out not to be feasible: nfc_data_generator is a public SDK API consumed by 5 external FAPs (nfc_magic, nfc_maker, flipcrypt, metroflip, seader), so moving it out of lib/ / the API would break their builds and force a major API bump. See the finding on #1033.

What's left is the safe, non-breaking part: trimming redundancy in place. Also removes 3 copies of the admin-key mapping that had to be kept in sync by hand.

Flash

Measured COMPACT=1 DEBUG=0, this branch vs dev:

Section	dev	this PR	Δ
.text	702520	702104	−416 B
.rodata	172452	172556	+104 B
.free_flash	172572	172884	+312 B free

~0.30 KiB reclaimed — modest, but free, plus the maintainability win of one source of truth.

Testing

• fbt firmware_all COMPACT=1 DEBUG=0 — links clean.
• fbt fap_nfc COMPACT=1 DEBUG=0 — APPCHK passes (NFC FAP still resolves the unchanged generator API).
• clang-format conformant.
• Reviewed as behavior-preserving: every one of the 18 generator rows verified against the original thunks (name/uid_len/type/size/ATS, in enum order); the listener's default fall-through semantics preserved exactly; poller authenticates the same addresses.

Refs #1033

[github-actions]

🤖 Build & analyze — refactor-mf-plus-generator-cleanup @ 75c15c74

✅ Firmware built — artifact unleashed-fw-refactor-mf-plus-generator-cleanup-75c15c74 (run)

💾 Flash & RAM — f7 (ble_light)

Metric	Size
Firmware (flash)	855.18 KiB
DFU image	855.48 KiB
Reserved — radio stack + FUS	164.00 KiB
Free flash (usable)	4.82 KiB ⚠️
RAM (.data + .bss)	7.94 KiB

_{1 MiB flash = firmware + usable free + radio/FUS. The BLE coprocessor stack sits at the top of flash (load addr 0x080D7000); the linker's .free_flash (168.82 KiB) counts that region as free, so usable free = .free_flash − reserved (164.00 KiB).}

📋 Public API changes

✅ No public API changes in this PR.