Explicit `text-to-video` pipeline tag

This PR explicitly sets `text-to-video` as pipeline_tag instead of getting it inferred by the server. It will help with discoverability + enable the text-to-video inference widget on the model page.

Note that the model is still tagged as both `text-to-video` and `image-to-video` so nothing will break in term of usage.

cc @sbrandeis

README.md

@@ -1,12 +1,12 @@
 ---
 tags:
 - ltx-video
-- text-to-video
 - image-to-video
 pinned: true
 language:
 - en
 license: other
+pipeline_tag: text-to-video
 ---
 
 # LTX-Video Model Card