fix: restore dynamic_fork_tasks_param in DynamicForkTask (re-fixes #349)

[nthmost-orkes]

Summary

Re-applies the one-line fix from PR #352, which was dropped from main when the branch was subsequently force-pushed/rebased (likely during the Agentspan consolidation). GitHub still shows #352 as "Merged" because it records the merge event, but commit a561d90 is not an ancestor of current main.

Actual timeline:

• baadd20 (Jan 2025) — PR FIX - Dynamic Fork (workflow as code)  #274 redesigned DynamicForkTask (replaced pre_fork_task with tasks_param/tasks_input_param_name) but set the wrong field: dynamic_fork_join_tasks_param instead of dynamic_fork_tasks_param. This is the bug.
• PR Updated dynamic for task to set params in right way #352 (Oct 2025) — Igor fixed the field name. Fix merged but lost when main was rewritten.
• This PR — Re-applies the same fix and adds unit tests to catch any future regression.

Setting dynamic_fork_join_tasks_param alongside dynamic_fork_tasks_input_param_name violates Conductor's validation: only one approach is valid at a time (dynamicForkJoinTasksParam alone, or dynamicForkTasksParam + dynamicForkTasksInputParamName together).

User impact: Users calling DynamicForkTask with separate tasks_param / tasks_input_param_name would get a 400 validation error from the server on workflow registration. Now registers correctly.

Test plan

• tests/unit/workflow/test_dynamic_fork_task.py — two new unit tests cover the correct param and the join-task variant
• Integration test generate_dynamic_fork_task in tests/integration/metadata/test_workflow_definition.py already exercises this path against a live server

Fixes #377

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
...onductor/client/workflow/task/dynamic_fork_task.py	100.00% <100.00%> (+100.00%)	⬆️

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mp-orkes]

The fix seems to be right but the description has the order of things backwards and CI is failing.

This PR #352 was undone by this commit: baadd20

This came first: #274
Then this: #352

A January commit can't revert an October PR. That PR was the fix. How did we "lose" the fix?

[mp-orkes]

LGTM - but please fix the description. We need to understand how this bug came back.

[nthmost-orkes]

The fix seems to be right but the description has the order of things backwards and CI is failing.

CI failure seems unrelated -- can look into it later today, posting in Slack channel for more visibility in case someone else can get to it first. (@chrishagglund-ship-it ?)