feat(tracing): add TypeScript SDK and publish pipeline

[davidfarah2003]

Summary

• Add TypeScript tracing SDK (@kayba_ai/tracing) wrapping mlflow-tracing — mirrors the Python SDK API
• Restructure tracing code into sdk/ directory with standalone kayba-tracing Python package
• ace.tracing re-exports from kayba_tracing so existing imports keep working
• Update CI publish workflow for all three packages with trusted publishing (OIDC)
• Bump all package versions to 0.9.5

SDK usage

Python (unchanged):

from ace.tracing import configure, trace
# or standalone:
from kayba_tracing import configure, trace

TypeScript (new):

import kayba from "@kayba_ai/tracing";
kayba.configure({ apiKey: "kb-...", folder: "my-project" });
const myFunc = kayba.trace(async () => { ... }, { name: "my_func" });

Test plan

• All 26 Python tracing tests pass with updated imports
• TypeScript SDK compiles and builds (CJS + ESM + .d.ts)
• Tested end-to-end against local Kayba backend with real LLM calls
• Verified traces stored correctly with full span tree in markdown
• Published @kayba_ai/tracing@0.9.4 to npm, verified install from fresh project

🤖 Generated with Claude Code