feat(config): add rope_temporal_size parameter

P1: support configurable temporal size for RoPE
- None (default): use actual frame count from input
- int value: fixed temporal size (e.g., 64 for legacy behavior)

Author: Luodian

onevision_encoder/configuration_onevision_encoder.py

@@ -77,6 +77,7 @@ def __init__(
         attention_dropout=0.0,
         initializer_range=0.02,
         rope_theta=10000.0,
+        rope_temporal_size=None,
         use_head=True,
         **kwargs,
     ):
@@ -94,4 +95,5 @@ def __init__(
         self.attention_dropout = attention_dropout
         self.initializer_range = initializer_range
         self.rope_theta = rope_theta
+        self.rope_temporal_size = rope_temporal_size  # None=use actual frames, int=fixed size (legacy: 64)
         self.use_head = use_head