[PT2.6][UX]Torch initialize and verbose message are clear and comparable with PyTorch GPUs

[riverliuintel]

🚀 The feature, motivation and pitch

PyTorch XPU verbose log should be clear and comparable with PyTorch GPU practice. Need to investigate and give the clear implementation requirement on different scenarios.
For example,

1. in initialize phase, Torch GPU should give clear software stack, running GPU information.
2. If run Torch XPU in a host with CPU only, it should give the CPU fallback information.
3. when run workload, it should give comparable verbose message as stock Pytorch.
4. when run in error, it should give helpful error message and reminder to report issue or get help message.
5. error message for some other exceptional cases

Alternatives

No response

Additional context

No response