Add Flora inference examples

Adds the TempSeg-to-Captioner inference examples from flora-share/Flora-TempSeg-Release-Infer@f495c8c709c72ca1fa39a71c02436834a4f6aebf. Binds the examples to immutable NVIDIA revisions with corrected model-card checksums. This PR and the corresponding nvidia/Flora-Captioner PR are a paired update.

The existing model card, weights, provenance, licenses, configuration, and templates are preserved. Only the model-card checksum entry changes among existing files. Static syntax, code-manifest, vendor-manifest, and model-asset identity checks passed; CUDA inference was not run. Keep the referenced PR commits available; if the merge strategy does not retain them, update both revision pins before deleting the PR refs.

SOURCE_MANIFEST.sha256

@@ -1,5 +1,5 @@
 837c0aa8748126d2193799e7bec51d6550ca4ba07767591e66dcb370c3115e44  MODEL_PROVENANCE.json
-6c051cfaeac3bbce138ba1b95ff1e21c7f2812356c472e9a11cf4b9467099680  README.md
+88cb6731fc521a0404faed7c5045c3db77185341849bc4e0b6e60be3443f7d6c  README.md
 ac4d8abd2a14045b22151354a7493049d651ec10c0b019fbe4927805e45c4b81  adapter_config.json
 1d80fe544e4cd1a7ff6b0ea4be71e988972408dab81d4c8af72e3a628739d6c3  adapter_model.safetensors
 d28ca68768ac2ce0ad42eaaf54d5bac4f8a298f7d017f5a04f3866094c65a839  chat_template.jinja