Eval bug: OuteTTS 0.3 and up crashes

[Som-anon]

Name and Version

~/work/llama.cpp/build/bin/llama-tts --version
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) UHD Graphics 620 (KBL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | warp size: 32 | shared memory: 65536 | int dot: 0 | matrix cores: none
version: 5072 (4ccea213)
built with cc (GCC) 14.2.1 20250207 for x86_64-pc-linux-gnu

So when using llama-tts to generate voice using OuteTTS 0.2 it works

`~/work/llama.cpp/build/bin/llama-tts -m /mnt/gguf/TTS/OuteTTS-0.2-500M-Q6_K.gguf -mv /mnt/gguf/TTS/WavTokenizer-Large-75-F16.gguf -p "Hello world"`
When trying the same thing using OuteTTS-0.3
`~/work/llama.cpp/build/bin/llama-tts -m /mnt/gguf/TTS/OuteTTS-0.3-1B-Q6_K.gguf -mv /mnt/gguf/TTS/WavTokenizer-Large-75-F16.gguf -p "Hello world"`
it crashes output attached as file. The same goes for OuteTTS-1.0

[outetss-0.3-fail.txt](https://github.com/user-attachments/files/19639189/outetss-0.3-fail.txt)


### Operating systems

Linux

### GGML backends

CPU, Vulkan

### Hardware

T480 i5-8250U 64G Ram

### Models

OuteTTS-0.3 (official GGUF) and WaveTokenizer (official GGUF)

### Problem description & steps to reproduce

* Get tts working with OuteTTS 0.2
* Try it switching only the OuteTTS 0.2 GGUF with the official 0.3 or 1.0 GGUF

### First Bad Commit

_No response_

### Relevant log output

```shell
~/work/llama.cpp/build/bin/llama-tts -m /mnt/gguf/TTS/OuteTTS-0.3-1B-Q6_K.gguf -mv /mnt/gguf/TTS/WavTokenizer-Large-75-F16.gguf -p "Hello world"`

[Som-anon]

sigh just saw #12794