Add /engines/unload

[doringeman]

On top of #45.

E.g.,

$ MODEL_RUNNER_HOST=http://localhost:8080 docker model run ai/smollm2 hi
$ MODEL_RUNNER_HOST=http://localhost:8080 docker model unload --all
Unloaded 1 model(s).
$ # run again
$ MODEL_RUNNER_HOST=http://localhost:8080 docker model unload ai/smollm2
Unloaded 1 model(s).
$ # run again
$ MODEL_RUNNER_HOST=http://localhost:8080 docker model unload --backend llama.cpp ai/smollm2
Unloaded 1 model(s).