fix: skip missing jinja2 variables instead of raising LLMNodeError

[EvanYao826]

Fixes langgenius/dify#38655

Summary

When a Jinja2 variable referenced in prompt_config.jinja2_variables is not present in the variable pool (e.g. because a conditional branch was not taken), _fetch_jinja_inputs raises VariableNotFoundError and the workflow fails — even though the Jinja2 template already handles optional variables with constructs like {% if variable is defined %}.

Root Cause

_fetch_jinja_inputs calls _get_required_variable which raises VariableNotFoundError when the variable pool returns None. The sibling code path _render_jinja2_message already handles this correctly by using variable_pool.get() and falling back to empty string.

Fix

Replace the _get_required_variable call in _fetch_jinja_inputs with a direct variable_pool.get() and skip variables that are not in the pool. This aligns with how _render_jinja2_message already handles missing variables.

Test plan

• Existing test test_fetch_jinja_inputs_raises_for_missing_variable in dify's test suite updated to expect the new (skip) behavior
• No regressions: _fetch_jinja_inputs is called during LLM node _run() — missing variables are simply omitted from the returned dict, which is used for input collection, not the Jinja2 rendering itself

[github-actions]

Thanks for the contribution. Before we can review or merge this pull request, please read our CLA and add this exact comment once: I have read the CLA Document and I hereby sign the CLA

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}