[BUGFIX] Fix intgemm flaky test in #19197 for master

[kpuatamazon]

Description

This is a twin of #19201 to fix #19197. I made a flaky test, sorry.

It's testing a quantizer that entails rounding but rounding 0.5 is ambiguous. So I intended to ban 0.5 by adding a small value. The code checks the interval (0.45, 0.55) and adds 0.05 to anything there. But that just shifts it to (0.5, 0.6) when I meant to go further from 0.5. Now it shifts to (0.65, 0.75) so 0.5 won't appear.

Checklist

Essentials

• PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage
• Code is well-documented

Changes

Fix flaky test test_contrib_intgemm_prepare_data

Comments

Another approach would be to allow some tolerance for quantized values around 0.5.

[mxnet-bot]

Hey @kpuatamazon , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

• To trigger all jobs: @mxnet-bot run ci [all]
• To trigger specific jobs: @mxnet-bot run ci [job1, job2]

---

CI supported jobs: [clang, website, unix-cpu, sanity, unix-gpu, windows-cpu, edge, centos-gpu, miscellaneous, windows-gpu, centos-cpu]

---

Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.