fix: run indexer reliably under pm2

[Snepsid]

Summary

• use a project-local PM2 dependency for the indexer engage scripts
• point PM2 at the actual Nest build outputs and align copied static assets with the runtime path
• log unhandled promise rejections before exiting so PM2 can restart with a useful error trail

Why

When the indexer was running through a long backlog, an unhandled promise rejection would terminate the process. The existing PM2 workflow also depended on a global PM2 install and pointed at build output paths that did not match the actual Nest artifact layout.

This change intentionally makes the existing engage, engage:mainnet, and engage:sepolia scripts self-contained and reproducible instead of relying on an undocumented global PM2 installation.

Dependency note

PM2 was added as a project dependency on purpose rather than assuming a global install.

Benefits:

• reproducible PM2 version across machines
• no pm2: not found setup failures on fresh environments
• the existing package scripts work without additional machine-specific setup
• easier debugging because the repo controls the PM2 version in use

The large yarn.lock diff is the expected result of adding PM2's transitive dependency tree, not unrelated lockfile churn.

Testing

• NETWORK=sepolia yarn build:sepolia
• started the sepolia app with PM2
• verified the process came online and resumed queueing blocks

[chopperdaddy]

what kind of temu chinese version of claude are you using to submit these PRs.

[Snepsid]

ah crap! I just realized I'm on Auto agent mode in copilot for this project. 💀

thanks for pointing that out. lol