feat: add initial eval scripts for unified memory

[maxervo]

Description

Eval scripts to get started on automating evaluating unified memory with scenarios/questions from body cam videos.
Iterated upon as we are running tests.

Drive the eval deterministically with context isolation for eval purity.

WIP in this draft:

• Code/folders being restructured in this draft
• Preparing instructions to run and reproduce results
• JSON standardization, better reporting etc.

Next steps after this PR:

• Concurrent tasks for independent run. Specify option for one run. Faster feedback
• Modularize code to cover both single/cross evals. Design abstractions and reduce duplicate code
• Better abstraction: decouple agent harness being tested so we can drive it with others (hermes, existing vss agent interactions)
• Refactor with existing VSS utilities, maybe NAT eval etc. (For now, just get it working with basics/simple python to get started and get initial results)

Checklist

• I am familiar with the Contributing Guidelines.
• I have installed and run pre-commit hooks locally (uv run pre-commit install once, then hooks run on every git commit).
• Every commit on this PR is DCO sign-off'd (git commit -s adds a Signed-off-by trailer that certifies you have the right to submit the change under Apache-2.0).
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[bp-sre-sonarqube]

Quality Gate passed for 'video-search-and-summarization-skills'

Issues
0 New issues
0 Fixed issues
0 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
No data about Duplication

See analysis details on SonarQube

[bp-sre-sonarqube]

Quality Gate passed for 'video-search-and-summarization-ui'

Issues
0 New issues
0 Fixed issues
0 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
No data about Duplication

See analysis details on SonarQube

[bp-sre-sonarqube]

Quality Gate passed for 'video-search-and-summarization-agent'

Issues
0 New issues
0 Fixed issues
0 Accepted issues

Measures
0 Security Hotspots
No data about Coverage
No data about Duplication

See analysis details on SonarQube