miner.py: implement checks and error handling for failing axons

[mvds00]

The miner template seemingly assumes that starting an axon never fails. In e2e tests (that use this template as their miner) the axon failed to start, due to mixing next_asyncio in bittensor and regular asyncio in uvicorn. It would be better to terminate the miner process when the axon never starts at all.

This patch addresses this by:

• wrapping run() in a try/except (this is a must in any Python threading application)
• signalling exceptions from worker to main thread in a thread safe manner
• terminating the miner if starting the axon fails
• monitoring and reporting on whether the axon still runs

Whether to keep the miner running if axon issues arise later is another question; the code as-is indicates this is indented behavior: "# In case of unforeseen errors, the miner will log the error and continue operations." so this is not changed.

This patch depends on another patch to bittensor that adds .is_running() and .exception to class axon.