Refactor command lookup per device

[monsivar]

Summary

Refactor command lookup so command classes can be resolved from the capabilities configured for a specific device, instead of relying only on the global command-name dictionaries.

This is a prerequisite for supporting different command implementations that use the same command name on different device families.

Motivation

This addresses the architectural blocker discussed in #1624.

The current global command lookup requires command names to be unique. That prevents, for example, registering two device-specific command classes that both use the same MQTT/API command name.

This PR adds a device-specific lookup while retaining the existing global registries as fallbacks for backwards compatibility.

Changes

Device capabilities

• Extract command classes configured in the capability tree.
• Store a per-device command lookup in Capabilities.
• Add Capabilities.get_command(name).
• Preserve the first/primary configured command when multiple capabilities use the same command name.
• Keep commands hidden behind lambdas or otherwise not directly discoverable working through the existing global fallback.

Legacy message lookup

• Pass the device's capabilities into JSON legacy message lookup.
• Resolve the corresponding command from the device first.
• Fall back to the existing global COMMANDS lookup when no device-specific command is available.

MQTT P2P lookup

• Resolve P2P command classes from the subscribed device's capabilities first.
• Use the receiver device for P2P requests (q).
• Use the sender device for P2P responses (p).
• Fall back to the existing global P2P command registry when the device does not configure the command.
• Do not fall back to another global implementation when the device explicitly configures the same command name with a non-P2P command.

Tests

Added coverage for:

• command lookup being specific to device capabilities
• different command classes sharing the same command name
• device-specific legacy message lookup
• device-specific MQTT P2P command lookup
• global MQTT P2P fallback
• preventing an incorrect global P2P fallback for an explicitly configured device command
• correct P2P device routing:
  • request → receiver device
  • response → sender device

Follow-up

Once this prerequisite is accepted, device families can safely select different command implementations with the same command name, including the mower clean command work discussed in #1624.

[codecov]

Codecov Report

❌ Patch coverage is 97.56098% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 96.33%. Comparing base (5453eba) to head (2d50596).

Files with missing lines	Patch %	Lines
deebot_client/messages/json/__init__.py	85.71%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##              dev    #1772      +/-   ##
==========================================
+ Coverage   96.26%   96.33%   +0.06%     
==========================================
  Files         161      161              
  Lines        6399     6434      +35     
  Branches      368      379      +11     
==========================================
+ Hits         6160     6198      +38     
+ Misses        172      170       -2     
+ Partials       67       66       -1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[codspeed]

Merging this PR will degrade performance by 30.42%

❌ 73 regressed benchmarks
✅ 153 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	test_onStationState_analyse[1-additional_content1]	5.2 ms	7.7 ms	-33.36%
❌	Simulation	test_onMapSetV2[199390082-vw]	5.2 ms	7.8 ms	-33.21%
❌	Simulation	test_onStationState_analyse[1-additional_content0]	5.2 ms	7.8 ms	-33.2%
❌	Simulation	test_onStationState_analyse[2-additional_content3]	5.2 ms	7.7 ms	-33.19%
❌	Simulation	test_onMapInfo_V2[2-5-False]	5.2 ms	7.7 ms	-33.17%
❌	Simulation	test_onStationState_analyse[1-additional_content2]	5.2 ms	7.7 ms	-33.12%
❌	Simulation	test_onMapInfo_V2[0-1-False]	5.2 ms	7.7 ms	-33.08%
❌	Simulation	test_onCachedMapInfo_no_using_map[0-expected_events1]	5.2 ms	7.8 ms	-32.85%
❌	Simulation	test_onAutoEmpty[False-None-None]	5.4 ms	8 ms	-32.68%
❌	Simulation	test_onMapSetV2[199390082-ar]	5.3 ms	7.9 ms	-32.59%
❌	Simulation	test_onGpsPos[139.691706-35.689487]	5.4 ms	8 ms	-32.51%
❌	Simulation	test_onAutoEmpty[False-10-10]	5.4 ms	8 ms	-32.49%
❌	Simulation	test_onAutoEmpty[True-10-10]	5.4 ms	8 ms	-32.49%
❌	Simulation	test_onStationState[1-additional_content1-1]	5.4 ms	8 ms	-32.48%
❌	Simulation	test_onMapSetV2[199390082-mw]	5.3 ms	7.9 ms	-32.47%
❌	Simulation	test_onAutoEmpty[False-auto-auto]	5.4 ms	8 ms	-32.46%
❌	Simulation	test_onAutoEmpty[False-smart-smart]	5.4 ms	8 ms	-32.45%
❌	Simulation	test_onStats[data1-expected1]	5.4 ms	7.9 ms	-32.44%
❌	Simulation	test_onMapInfo_V2[1-1-True]	5.4 ms	8 ms	-32.43%
❌	Simulation	test_onAutoEmpty[True-auto-auto]	5.4 ms	8 ms	-32.39%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing monsivar:refactor/device-command-lookup (2d50596) with dev (5453eba)}

[monsivar]

@edenhaus This PR addresses the per-device command lookup prerequisite you mentioned in #1624.

All CI checks are passing. When you have time, I’d appreciate a review.