Benchmarks: Micro benchmark - Add simultanneously all-to-host / host-to-all bandwidth testcases to nvbandwidth

[yukirora]

Description
Add simultanneously all-to-host / host-to-all bandwidth testcases to nvbandwidth .

Major Revision

• nvbandwidth.patch: Add simultanneously all-to-host / host-to-all bandwidth testcases to nvbandwidth
• upgrade nvbandwidth submodule into v0.8
• add patch into makefile build

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.97%. Comparing base (76066b6) to head (bfc2d7b).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #736   +/-   ##
=======================================
  Coverage   85.97%   85.97%           
=======================================
  Files         102      102           
  Lines        7578     7578           
=======================================
  Hits         6515     6515           
  Misses       1063     1063           

Flag	Coverage Δ
cpu-python3.10-unit-test	71.05% <ø> (ø)
cpu-python3.12-unit-test	71.05% <ø> (ø)
cpu-python3.7-unit-test	70.49% <ø> (ø)
cuda-unit-test	83.86% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[yzygitzh]

A general question: why do we need to test fully parallel all-to-host and host-to-all?

[yukirora]

A general question: why do we need to test fully parallel all-to-host and host-to-all?

we want to check the nvidia c2c performance between grace gpu and grace cpu, which need fully parallel alltohost and sum, if we only use one gpu to the cpu, it's not accurate

[guoshzhao]

We have another PR [#739] will do the nvbandwidth submodule upgrade, let's see which one can be the first to merge, then we can remove the upgrade from the other PR.