feat(ai): add LangChain v1 agent middleware (#4556)

* feat(ai): add LangChain v1 agent middleware

* fix(ai): preserve middleware trace hierarchy

* chore(ai): prepare LangChain middleware for latest main

Refresh the AI package manifest and lockfile with the dependency updates already on main so GitHub can update the PR branch without conflicts.

* chore(ai): avoid generated lockfile merge overlap

Keep the LangChain-only lockfile entries together so the current main lockfile can merge without content conflicts. This changes ordering only.

* fix(ai): align LangChain middleware with agent requests

Preserve explicitly declared custom state, normalize model and tool metadata, and capture the messages LangChain actually sends.

* fix(ai): omit dangling LangChain parent IDs

Keep terminal model and tool failures on their trace without pointing to a root that LangChain never emits. Cover propagated agent latency and retry recovery.

---------

Co-authored-by: Manoel Aranda Neto <5731772+marandaneto@users.noreply.github.com>