[Feature-Request] : add --skip_seconds argument

[lightsabata]

Hi everyone! I know it is kind of a "niche" feature, but would it be possible to add a kind of --skip_seconds argument so the sentence registered is ignoring the first miliseconds of the audio.

To explain : Before in rhasspy and now with homeassistant-satellite (which is much more convinent) i set into my home a kind of a virtual butler (Alfred, from Batman thanks to porcupine custom wake word). In this scenario, a wake sound is enabled and plays an audio file wich is in fact a tts who says "monsieur ?" ("sir ?" but in french). Obvously this kind of setting mess the intend recognition in many ways.

Even with a noise cancelling parameter and with skip_word added in on Home Assistant, the wake word sound is many times registered as part of the the intend and/or mess the vad process.

On rhasspy 2.5 i use to have the skip_seconds parameter enabled on my config file.

I don't know if it should be added on the satellite main part or on the vad side, i have tried to made some tests based on rhasspy 2.5 but it is way far form my poor python skills.

Thanks