fix(assets): resolve assets by displayName when no PID matches (MX Master 3S over BTLE)

[jcbrizu]

Problem

On an MX Master 3S connected through a Logi Bolt receiver, OpenLogi shows no device image — the GUI falls back to the synthetic silhouette and the asset depot is never downloaded (sync: no matching depots for connected devices).

Root cause: over BTLE the device's HID++ DeviceInformation (feature 0x0003) reports model_ids = [0xb034], extended_model_id = 1, i.e. config_key = "1b034". The asset registry, however, keys the MX Master 3S as 2b043 (a different transport's PID). resolve_in_index only matches by PID — strict {ext}{pid} and bare-PID suffix — and the device never reports b043 under any transport, so nothing matches.

Observed with openlogi list on hardware:

slot 2 ● MX Master 3S (mouse, wpid=b034)
        model_ids=[b034,0000,0000] ext=01 transports=btle

Fix

Add a final resolution step that bridges the firmware codename to the registry displayName (exact, case-insensitive). "MX Master 3S" lines up even when no PID does. This generalises to any BTLE-connected device whose live PID is absent from the registry.

The codename is threaded through resolve_in_index, AssetResolver::resolve, and the asset sync path (collect_models now carries (DeviceModelInfo, Option<String>)). PID matching still takes precedence; the name match is a last resort, so existing devices are unaffected.

Verification

• cargo test -p openlogi-assets -p openlogi-gui — green, including new tests for Index::find_by_display_name and for the codename fallback in resolve_in_index (a BTLE 3S model misses by PID, resolves by codename).
• On hardware (MX Master 3S / Bolt / macOS 26): the log now shows asset matched via codename↔displayName fallback depot="mx_master_3s", the mx_master_3s depot downloads, and the mouse image renders.

Likely also helps the missing-image symptom reported in #19 / #25.

[pullfrog]

✅ No new issues found.

Reviewed changes — adds a codename ↔ displayName last-resort tier to the asset resolver so an MX Master 3S on BTLE (and similar devices whose live PID isn't in the registry under any transport) still picks up its depot.

• Add Index::find_by_display_name — exact, case-insensitive lookup over the registry's displayName. Unit tests cover both case-insensitivity and "exact, not substring" (so MX Master 3 cannot bleed into the MX Master 3S depot).
• Thread codename through AssetResolver::resolve and resolve_in_index — match order documented in the rustdoc as force → strict PID → suffix PID → codename. New module tests demonstrate the bug (PID-only misses BTLE 3S) and the fix (codename bridges to the mx_master_3s depot).
• Propagate the codename through the sync path — sync::sync now takes &[(DeviceModelInfo, Option<String>)], collect_models pairs model_info with paired.codename, and the GUI render call site passes paired.codename.as_deref() to cache.resolve.

PID matching still takes precedence, the name match runs only after both PID tiers miss, and the codename comes from firmware (Bolt receiver) or a hard-coded synthetic lookup — never from user input — so the new tier doesn't widen the matching surface in a risky way. The fallback chain inside load_files (variant → side_core.png → front_core.png) already handles the case where a codename-matched depot's manifest.json doesn't have an entry for the device's reported extended_model_id, so a depot hit still renders something.

  ^{｜ View workflow run ｜ Using Claude Opus ｜ 𝕏}

[AprilNEA]

Rebased onto master (dropping the merge commit) and landed directly:

• 7505fa5 — your fix, authored by you
• f64f40c — a one-line serial_number: None follow-up the btle_3s_model test helper needed after master added that field to DeviceModelInfo (otherwise the merged tree didn't compile)

Thanks @jcbrizu! Closing since the change is now on master.