Add CFG condition for batch size adjustments

[xenova]

Currently, cfg_weight=0.0 fails because the current inference code assumes we always perform classifier free guidance. As a result, the following code fails:

import torchaudio as ta
from chatterbox.tts import ChatterboxTTS

model = ChatterboxTTS.from_pretrained(device="cpu")

text = "Ezreal and Jinx teamed up with Ahri, Yasuo, and Teemo to take down the enemy's Nexus in an epic late-game pentakill."
wav = model.generate(text, cfg_weight=0.0)
ta.save("test-1.wav", wav, model.sr)

with this error:

    wav = model.generate(text, cfg_weight=0.0)
  File "/Library/Frameworks/Python.framework/Versions/3.13/lib/python3.13/site-packages/chatterbox/tts.py", line 246, in generate
    speech_tokens = self.t3.inference(
        t3_cond=self.conds.t3,
    ...<6 lines>...
        top_p=top_p,
    )
  File "/Library/Frameworks/Python.framework/Versions/3.13/lib/python3.13/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/Library/Frameworks/Python.framework/Versions/3.13/lib/python3.13/site-packages/chatterbox/models/t3/t3.py", line 309, in inference
    inputs_embeds = torch.cat([embeds, bos_embed], dim=1)
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 1 but got size 2 for tensor number 1 in the list.

---

This PR fixes this, allowing cfg_weight to be set to 0. It's ~2x faster, but does produce slightly worse results.