Log warmup overhead time

[adarshyoga]

• Have you provided a meaningful PR description?

Adds an entry to the output log called warmup overhead time which is the difference of the median execution time and warm up time. This ovehead time points to the time spent in jit-compilation (as in the case of numba-dpex) or framework load time (as in the case of dpnp, numpy). Sample output below.

Fixes #336 .

---

================ implementation numba_dpex_k ========================
implementation: numba_dpex_k
framework: numba_dpex
framework version: 0.22.0.dev2+3.g59d523892
input size: 20971520
setup time: 372.803065ms (372803065 ns)
warmup time: 821.023662ms (821023662 ns)
teardown time: 4.727934ms (4727934 ns)
max execution times: 2.488874ms (2488874 ns)
min execution times: 2.062158ms (2062158 ns)
median execution times: 2.091041ms (2091041 ns)
warmup overhead time (warmup time - median execution time): 818.932621ms (818932621 ns)
repeats: 10
preset: S
validated: Success

• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• If this PR is a work in progress, are you filing the PR as a draft?