fix: remove redundant autoconf from LCMService.start()

[SUMMERxYANG]

Problem

After PR #1503 (mac configurator) merged, users see repeated "Apply these changes now? [y/N]" prompts on startup — once per LCMService instance across all worker processes. The system configuration prompt loops indefinitely.

Solution

Remove the autoconf() call from LCMService.start(). The blueprint system already runs all system configurators via _run_configurators before deploying any modules to workers, making the per-service call redundant. Standalone scripts (e.g. drone.py) also call autoconf() explicitly before creating any services.

Also removes the now-dead LCMConfig.autoconf field and cleans up all callsites that passed it, since the field was only ever read by the deleted code path.

Breaking Changes

• LCMConfig.autoconf field removed. Any code passing autoconf=True/False to LCMService, LCMPubSubBase, LCM, PickleLCM, etc. will get a TypeError. Fix: remove the autoconf= kwarg.

How to Test

1. Run a blueprint that spawns multiple workers (e.g. G1 agentic blueprint)
2. Verify the system configuration prompt appears only once on startup
3. Verify LCM communication still works correctly

Contributor License Agreement

• I have read and approved the CLA.

[greptile-apps]

Greptile Summary

This PR fixes a regression introduced in #1503 where the autoconf() call inside LCMService.start() caused repeated interactive "Apply these changes now? [y/N]" prompts — once per LCM service instance across every worker process. The fix removes autoconf() from LCMService.start(), removes the now-dead LCMConfig.autoconf field, and cleans up all 14 call sites that were passing the autoconf= kwarg. The autoconf() function itself is preserved in lcmservice.py and re-exported via lcmpubsub.__all__, so standalone scripts (like drone.py) that call it explicitly before spinning up services continue to work correctly.

• LCMConfig.autoconf field removed — a breaking change: any external code still passing autoconf=True/False to LCMService, LCMPubSubBase, LCM, PickleLCM, etc. will receive a TypeError.
• All 14 in-tree call sites updated correctly.
• Three test methods that validated the old autoconf= behavior are properly deleted.
• docs/usage/transports/index.md:247 and docs/development/grid_testing.md:84 still show LCM(autoconf=True) in code snippets and were not updated — users following either document will hit an immediate TypeError.

Confidence Score: 4/5

• Safe to merge — the logic change is correct and all in-tree code is updated, but two documentation files still reference the removed autoconf=True kwarg.
• The core fix is clean and well-reasoned: removing a redundant per-instance autoconf() call that was causing repeated interactive prompts across worker processes. All 14 production and test call sites are correctly updated. The only gap is that two docs files (docs/usage/transports/index.md and docs/development/grid_testing.md) were not updated and still show the now-invalid LCM(autoconf=True) pattern, which will cause a TypeError for any user following those guides.
• docs/usage/transports/index.md (line 247) and docs/development/grid_testing.md (line 84) still reference the removed autoconf=True kwarg.

Important Files Changed

Filename	Overview
dimos/protocol/service/lcmservice.py	Core fix: removes autoconf() call from LCMService.start() and drops the LCMConfig.autoconf field; clean, correct change that eliminates the redundant per-worker prompt.
dimos/protocol/service/test_lcmservice.py	Tests properly updated: removes the three autoconf-specific test cases and the patch("autoconf") wrapper from all remaining service tests.
dimos/protocol/pubsub/impl/test_lcmpubsub.py	Fixtures updated to drop autoconf=True from LCMPubSubBase, PickleLCM, and LCM constructors.
dimos/utils/cli/dtop.py	Removes autoconf=True from PickleLCM constructor; the explanatory comment about TUI deadlocking is retained, which is still useful context.
dimos/utils/cli/lcmspy/run_lcmspy.py	Drops autoconf=True from GraphLCMSpy; comment about TUI deadlock is preserved for context.
dimos/visualization/rerun/bridge.py	Removes autoconf=True from all three LCM() instantiations in config defaults, docstring example, and run_bridge.
dimos/robot/unitree/g1/blueprints/primitive/uintree_g1_primitive_no_nav.py	Removes autoconf=True from the G1 blueprint's rerun_config LCM instantiation.
dimos/robot/unitree/go2/blueprints/basic/unitree_go2_basic.py	Removes autoconf=True from the Go2 blueprint's rerun_config LCM instantiation.

Sequence Diagram

sequenceDiagram
    participant BP as Blueprint / standalone script
    participant RC as _run_configurators
    participant LS as LCMService.start()
    participant AC as autoconf()

    Note over BP,AC: Before this PR (broken)
    BP->>RC: _run_configurators() — once
    RC->>AC: autoconf() ✅ prompt shown once
    BP->>LS: worker 1: service.start()
    LS->>AC: autoconf() ❌ prompt shown again
    BP->>LS: worker 2: service.start()
    LS->>AC: autoconf() ❌ prompt shown again

    Note over BP,AC: After this PR (fixed)
    BP->>RC: _run_configurators() — once
    RC->>AC: autoconf() ✅ prompt shown once
    BP->>LS: worker 1: service.start()
    Note right of LS: no autoconf() call
    BP->>LS: worker 2: service.start()
    Note right of LS: no autoconf() call

Loading

Comments Outside Diff (1)

1. docs/usage/transports/index.md, line 247 (link)

   Stale autoconf=True kwarg in documentation

   This code snippet still uses LCM(autoconf=True), which is now removed as a breaking change in this PR. Any user following this documentation will get a TypeError: __init__() got an unexpected keyword argument 'autoconf' immediately.

   The same stale usage also appears in docs/development/grid_testing.md:84.

_{Last reviewed commit: d7d0580}

[SUMMERxYANG]

@greptile