[UX] minor optimizations for launch and introduce py-spy

[aylei]

I've investigated python profiling tools when developing #3159 since python -m cProfile only sees profile on main thread and profiling multi-thread applications using cProfile is tedious. I evaluated yappi and py-spy, it turns out the latter one has better UX and broader adoption. Therefore, I propose changing the profiling approach in dev guide to py-spy.

Also, I found some minor optimization chances and made some quicks in this PR. sky launch speed ​​up 100ms in average:

# PR branch, 26.16s
/usr/bin/time bash -c 'for i in {1..10}; do SKYPILOT_DISABLE_USAGE_COLLECTION=1 python -m sky.cli launch --gpus H100:8> /dev/null; done;'
       26,16 real        16,37 user        25,22 sys

# master branch, 27.26s
/usr/bin/time bash -c 'for i in {1..10}; do SKYPILOT_DISABLE_USAGE_COLLECTION=1 python -m sky.cli launch --gpus H100:8> /dev/null; done;'
       27,26 real        16,26 user        25,34 sys

Tested (run the relevant ones):

• Code formatting: bash format.sh
• Any manual or new tests for this PR (see benchmark above)
• All smoke tests: pytest tests/test_smoke.py
• Relevant individual smoke tests: pytest tests/test_smoke.py::test_fill_in_the_name
• Backward compatibility tests: conda deactivate; bash -i tests/backward_compatibility_tests.sh

appendix:

• GIL flamegrpah captured by py-spy

• Speedscope flamegraph captured by py-sy, which is more user-friendly

[aylei]

@Michaelvll please kindly review this PR