macOS compatibility

Goal: Get simulation, model inferencing, and training working on macbook

What I did:
- Refactored mujoco teleoperation interface to use flask web server
- mainly to circumvent macos quirks, but also future proofing for headless setup
- Got headless mujoco setup working
- refactored training script for headless environments
- Added model training resumption from weight file
- Stressed over grad school applications (will I get in anywhere that does VLA work???)
- Collected 150 more episodes of training data by supervising autonomous operation
- Allowed model to run in simulation, discarded failed episodes, saved successful ones
- surprisingly did not noticeably improve task performance after retraining

What worked:
- Was able to successfully get data collection and model testing working on mac, both locally and headless connection to pc
-
- Quality of life upgrades (cleaning up training script and adding training resumption) worked seamlessly

What failed:
- running mujoco locally on mac is inconvenient and not as well supported as linux
- tripling amount of collected episodes did not noticeably improve performance

Key learning:
- Macos and linux dont play as nicely together as i would have thought
- may need to start looking into docker when deploying on real cloud instances

Next session:
- begin sim 2 real testing and data collection
- write model inferencing code for ros2 / hardware interface
- inference on gpu, either write web server or use ros2 directly
- test pure sim policy on real hardware
- set up real world data collection pipeline (ros2 bag post-processing?)
- publish devlog on eryk.ca eventually

Time spent: 3 hour