[TVMC] End to end benchmarking by default

[tmoreau89]

End to end benchmarking was added to the VM and graph executor in order to faithfully represent execution time by accounting for data transfer overheads. This can be particularly significant on discrete GPUs, where PCI-E transfers are important to account for in a typical model serving deployment.

This PR proposes to modify the default measurement done by TVMC to always benchmark end to end execution time from CPU local device context.
Another option is to expose in TVMC a flag that lets the user perform end to end benchmarking. I recommend however not benchmarking without data transfer overheads as it presents an overly optimistic outlook on TVM performance.

[tmoreau89]

CC @mbrookhart @jwfromm @tkonolige @leandron

[jwfromm]

Thanks for making this change @tmoreau89. I think changing the default behavior is likely going to cause issues for entrenched users. I just put up PR #10256 which exposes an end_to_end argument to the python and command line interface and defaults it to False. I think this is a less disruptive solution.

[tmoreau89]

@jwfromm great points, I'll close this PR in favor of your PR which seems less disruptive