Flaky e2e: test_go_to_the_bed times out without reporting final position

dimos/e2e_tests/test_dimsim_spatial_memory.py::test_go_to_the_bed (self-hosted-large job) flakes with:

TimeoutError: Failed to get to position x=-3.567, y=-1.332

Example failure: https://github.com/dimensionalOS/dimos/actions/runs/30356553452/job/90265993009

The message reports the target, not where the robot ended up. lcm_spy.py:159 wait_until_odom_position computes distance on every odom message but throws away the last-seen position, so every flake is undebuggable.

Proposed fix:

1. Include the last odom position and the minimum distance reached in the TimeoutError message.
2. Evaluate a soft-failure margin: nav threshold is 2m/180s — decide whether ending within some larger margin of the bed (e.g. ≤3m) should warn/soft-pass instead of hard-fail.

Synced from DIM-1359 by stash

Start with dimos/e2e_tests/test_dimsim_spatial_memory.py::test_go_to_the_bed and lcm_spy.py:159, where wait_until_odom_position calculates distance and raises the timeout. Reproduce or inspect the self-hosted-large failure, then make the error report the last odom position and minimum distance reached. Resolve and test the proposed soft-failure margin for results near the bed, with the e2e test passing and failures remaining diagnosable.