Improvements to `lf.eval` framework.

[copybara-service[bot]]

Improvements to lf.eval framework.

• Add self-link for each task in summary.html. This allows easy navigation of a particular task among many tasks.
• Disable input function for LLM generated code, preventing the evaluation process from being blocked.
• Fix a prompt schema update in the Colab.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 96.00%. Comparing base (f049b81) to head (6f4b7ac).

:exclamation: Current head 6f4b7ac differs from pull request most recent head 7d728fa. Consider uploading reports for the commit 7d728fa to get more accurate results

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #163 +/- ## ======================================= Coverage 95.99% 96.00% ======================================= Files 55 55 Lines 4445 4450 +5 Branches 961 962 +1 ======================================= + Hits 4267 4272 +5 Misses 90 90 Partials 88 88 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.