feat(loging): Improve logging output of provider and get

[flub]

This makes the logging output a lot better, driven by real-life
debugging. It introduces a few more spans and downgrades the
per-frame stuff to trace.

[flub]

/netsim

[github-actions]

flub/improve-logging.80919111d422ab913d54a47151fe7c57da881a73
Perf report:

test	case	throughput_gbps
iroh_latency_200ms	1_to_1	2.46
iroh_latency_200ms	1_to_3	8.38
iroh_latency_200ms	1_to_5	10.06
iroh_latency_200ms	1_to_10	8.49
iroh_latency_200ms	2_to_2	5.89
iroh_latency_200ms	2_to_4	11.70
iroh_latency_200ms	2_to_6	15.31
iroh_latency_200ms	2_to_10	15.21
iroh_latency_20ms	1_to_1	2.63
iroh_latency_20ms	1_to_3	8.11
iroh_latency_20ms	1_to_5	9.77
iroh_latency_20ms	1_to_10	7.99
iroh_latency_20ms	2_to_2	5.59
iroh_latency_20ms	2_to_4	11.69
iroh_latency_20ms	2_to_6	15.34
iroh_latency_20ms	2_to_10	16.42
iroh	1_to_1	3.60
iroh	1_to_3	8.86
iroh	1_to_5	11.35
iroh	1_to_10	9.67
iroh	2_to_2	6.34
iroh	2_to_4	12.30
iroh	2_to_6	15.90
iroh	2_to_10	17.97

[b5]

can confirm output of provide is much nicer with this branch, especially with large directory trees