Latency trackers and latency reduction optimizations

[AuroraH456]

• Adds time duration printout for each stage of processing
• Before sending the screenshot to OpenAI for screen analysis, this will convert the screenshot from PNG to JPEG and decrease the quality. This decreases latency by 2-4 seconds (out of the original 10-ish) as most of the latency associated with using OpenAI API to analyze the screenshot comes from sending an image over to the OpenAI servers (this fact was found out experimentally through using different models of GPT).
• Uses a smaller model of GPT (gpt-4.1-nano instead of gpt-4o) for faster speed (screen analysis is fairly simple and shouldn't require a big model). However, this doesn't reduce latency by much as most of the processing time is be spent on sending the screenshot to OpenAI.

[ecoArcGaming]

lgtm 🔥

[AuroraH456]

Changed OpenAI TTS to Piper TTS, which decreases latency by 1-2 seconds. Piper TTS needs to be downloaded and run locally on the server; the download command is in the README file. The TTS models should be downloaded in the voice_response folder.