docs: model benchmarks (#17036)

* model benchmarks

* minor fixes

* formatting

* docs build

* maybe fix reference

* clarify optimal

* use emojis

* wording

* wording

* clarify optimal wording

* bolding

* more detailed instructions

* clarify edge case fix

* early exit in dim loop

machine-learning/.gitignore

@@ -12,6 +12,9 @@ onnx__*
 *.pos_embed
 vocab.txt
 export/immich_model_exporter/models/**/README.md
+export/**/results/*.json
+export/**/root
+*.armnn
 tokenizer.json
 tokenizer_config.json
 special_tokens_map.json