ollama can discard posts

Ollama, by default, can queue up to 512 requests (OLLAMA_MAX_QUEUE). If the queue is full, new requests are discarded. Hivesense, during synchronization, aggressively pushes requests as soon as it receives a new batch of blocks to process. It does not check whether Ollama can process or discard them. As a result, some requests may never be vectorized, making them unavailable for subsequent searches.