fix: pagination and tracing

[lc-arjun]

This PR does a few things:

1. fixes the server to actually support the x-pagination-total header
2. updates assistants search to return this header (already was included in threads search but not assistants)
3. includes flag for if langsmith tracing is enabled

all these are supported in the python server

[hinthornw]

For in-mem server wouldn't it be opt-in?

[lc-arjun]

this logic is the same as how we do it in python server. tracing is only disabled if you explicitly set langsmith_tracing or any of the variants to false or if you dont have an api key defined

[lc-arjun]

https://github.com/langchain-ai/langgraph-api/blob/main/api/langgraph_api/config.py#L328

[lc-arjun]

https://github.com/langchain-ai/langgraph-api/blob/00b0b34bf1f9ce8702e405ce36969464e8745794/api/langgraph_api/api/meta.py#L23

[hinthornw]

That's opt-in:

TRACING = (
    env("LANGCHAIN_TRACING_V2", cast=bool, default=None)
    or env("LANGCHAIN_TRACING", cast=bool, default=None)
    or env("LANGSMITH_TRACING_V2", cast=bool, default=None)
    or env("LANGSMITH_TRACING", cast=bool, default=None)
)

[hinthornw]

unless i'm misreading (which happens a lot)

[lc-arjun]

this logic is admittedly a bit hard to reason about but in the block above that we do:

if (
    LANGSMITH_API_KEY
    and not getenv("LANGCHAIN_TRACING_V2")
    and not getenv("LANGCHAIN_TRACING")
    and not getenv("LANGSMITH_TRACING_V2")
    and not getenv("LANGSMITH_TRACING")
):
    environ["LANGCHAIN_TRACING_V2"] = "true"

so if they have not set a value for langsmith_tracing but have an api key, it gets set to true