fix(BA-5986): make ModelMountConfigInput.definition_path nullable

Server already falls back to `model-definition.yaml`/`.yml` when
`model_definition_path` is unset, but the GQL/DTO boundary forced a
non-null value — so frontends sent a placeholder and the fallback never
ran when the actual file used the `.yml` extension. Relax the v2 DTO,
legacy DTO, and GQL input so omitting the field triggers auto-detection.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: seedspirit

docs/manager/graphql-reference/supergraph.graphql

@@ -9558,7 +9558,7 @@ input ModelMountConfigInput
 {
   vfolderId: ID!
   mountDestination: String!
-  definitionPath: String!
+  definitionPath: String = null
 }
 
 """

docs/manager/graphql-reference/v2-schema.graphql

@@ -6337,7 +6337,7 @@ type ModelMountConfig {
 input ModelMountConfigInput {
   vfolderId: ID!
   mountDestination: String!
-  definitionPath: String!
+  definitionPath: String = null
 }
 
 """

src/ai/backend/common/dto/manager/deployment/request.py

@@ -238,7 +238,14 @@ class ModelMountConfigInput(BaseRequestModel):
 
     vfolder_id: VFolderUUID = Field(description="Model vfolder ID")
     mount_destination: str = Field(default="/models", description="Mount destination path")
-    definition_path: str = Field(description="Model definition file path within vfolder")
+    definition_path: str | None = Field(
+        default=None,
+        description=(
+            "Optional model definition file path within vfolder. "
+            "When omitted, the server auto-detects `model-definition.yaml` or "
+            "`model-definition.yml`."
+        ),
+    )
 
 
 class ModelRuntimeConfigInput(BaseRequestModel):

src/ai/backend/common/dto/manager/v2/deployment/request.py

@@ -202,7 +202,14 @@ class ModelMountConfigInput(BaseRequestModel):
 
     vfolder_id: VFolderUUID = Field(description="VFolder ID for the model")
     mount_destination: str = Field(description="Mount destination path inside container")
-    definition_path: str = Field(description="Path to model definition file")
+    definition_path: str | None = Field(
+        default=None,
+        description=(
+            "Optional path to the model definition file within the model vfolder. "
+            "When omitted, the server auto-detects `model-definition.yaml` or "
+            "`model-definition.yml`."
+        ),
+    )
 
 
 class ExtraVFolderMountInput(BaseRequestModel):

src/ai/backend/manager/api/gql/deployment/types/revision.py

@@ -776,7 +776,7 @@ class ModelRuntimeConfigInput(PydanticInputMixin[ModelRuntimeConfigInputDTO]):
 class ModelMountConfigInput(PydanticInputMixin[ModelMountConfigInputDTO]):
     vfolder_id: ID
     mount_destination: str
-    definition_path: str
+    definition_path: str | None = None
 
 
 @gql_pydantic_input(