perf: Use Granian as app server

Much more performant and secure, made in Rust.

doc: Specify App Config refresh behavior

ux: Lower thresholds for timeout static prompt triggers

doc: Add related projects

chore: Remove dead code

fix: OTEL spans in async context

fix: Context windows for gpt-4.1 models family

quality: Code quality

perf: Migrate to gpt-4.1 and gpt-4.1-nano

Lower the answer latency by a few 10s or milliseconds.

fix: Add enough LLM quota for basic usage

150k tokens /sec