chore(llmobs): support running experiment tasks

[Yun-Kim]

MLOB-3272

Adds support for running experiment task functions concurrently, and tracing said experiment tasks.
Also does the following:

• Make DatasetRecord.input_data a more strict Dict[str, NonNoneJSONType]. There was a discrepancy between previous iterations of input_data from the DatasetRecord class and the required signature for task functions.
• Makes config a required input param (can be optional though) for task functions. Making it possibly optional is too clunky of an experience and users should be required to set it
• Private helper to create experiment spans to trace the experiment task run. Note that this is still submitted to the regular LLMObs span track, future PR will handle custom scoping and experiment custom I/O annotation on said span.

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_3bf4897d-e6aa-43a3-8d9c-5097b1f85177_batch_update_post_2d58a82a.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_67c7b6cc-ce98-481e-ab9b-e4925564826c_batch_update_post_a1f44751.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_delete_post_264c9f32.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_delete_post_d28fa230.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_post_35e295b1.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/datadog/datadog_api_unstable_llm-obs_v1_datasets_post_3f098e52.yaml  @DataDog/ml-observability
ddtrace/llmobs/_constants.py                                            @DataDog/ml-observability
ddtrace/llmobs/_experiment.py                                           @DataDog/ml-observability
ddtrace/llmobs/_llmobs.py                                               @DataDog/ml-observability
ddtrace/llmobs/_writer.py                                               @DataDog/ml-observability
tests/llmobs/test_experiments.py                                        @DataDog/ml-observability

[github-actions]

Bootstrap import analysis

Comparison of import times between this PR and base.

Summary

The average import time from this PR is: 280 ± 4 ms.

The average import time from base is: 282 ± 4 ms.

The import time difference between this PR and base is: -2.0 ± 0.2 ms.

Import time breakdown

The following import paths have shrunk:

ddtrace.auto 1.984 ms (0.71%)

ddtrace.bootstrap.sitecustomize 1.305 ms (0.47%)

ddtrace.bootstrap.preload 1.305 ms (0.47%)

ddtrace.internal.remoteconfig.client 0.658 ms (0.24%)

ddtrace 0.679 ms (0.24%)

ddtrace.internal._unpatched 0.032 ms (0.01%)

json 0.032 ms (0.01%)

json.decoder 0.032 ms (0.01%)

re 0.032 ms (0.01%)

enum 0.032 ms (0.01%)

types 0.032 ms (0.01%)

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-07-15 01:44:55

Comparing candidate commit c70bb0d in PR branch yunkim/dne-run-tasks with baseline commit c8d1431 in branch main.

Found 0 performance improvements and 2 performance regressions! Performance is the same for 546 metrics, 2 unstable metrics.

scenario:iastaspects-format_map_aspect

• 🟥 execution_time [+470.593ns; +565.544ns] or [+14.604%; +17.551%]

scenario:iastaspectsospath-ospathjoin_aspect

• 🟥 execution_time [+809.319ns; +996.221ns] or [+13.167%; +16.208%]

[Kyle-Verhoog]

some thoughts, i think i'd change the log/raise condition check for sure but trust you to address what you feel is necessary