Add encoder model support in trt-llm (#1294)

* add encoder

* update: port

* update: serving image builder

* fmt

* fmt

* fmt

* add variables for encoder

* update default config settings

* update: constants, serving image builder

* update: serving image builder

* update start command

* update truss serving builder

* add clearer separation between encoder and decoder paths

* remove concurrenct overwrite

* serving image builder: update predict concurrency

* add: comments for max batch size

* make serving image builder asserion

* add predict endpopint as /v1/embeddings for faster testing