Create worker data_dir in edge install scripts

[ajayshekar-s1]

Summary

1. Create worker data_dir (original commit)

• Every edge-config.json sets data_dir to $ROOT_DIR/data, and the Vector-based worker's config validation hard-fails with exit 78 (data_dir "..." does not exist) if that directory isn't already present — the worker does not create it itself.
• install.sh, install_darwin.sh, and install.ps1 only ever created the root, logs, and update directories, so every fresh install left the worker permanently unable to start, independent of any config content/entitlement issues.
• Adds DATA_DIR/$DataDir (overridable via env var, consistent with the other paths) and creates/chowns/chmods it alongside the existing directories in all three installers.
• This diff was intentionally scoped to only the data_dir addition — it preserves the edge_manager_tls_enabled fix from Sync edge install/uninstall scripts from pipeline/edge/configs/ #20 as-is.

2. Always set edge_manager_tls_enabled to true (added commit)

• Replaces the URL-scheme-based derivation (wss://|https:// → true, else false) with an unconditional true in all three install scripts.
• Every supported deployment terminates TLS at the edge-manager ingress, so the derivation was redundant in practice.
• Real behavior change, explicitly requested: a plaintext-only edge-manager deployment (INSECURE_PLAINTEXT_OPAMP=true, ws://) will now have its enrollment HTTP client unconditionally attempt https://, which has no fallback/retry and fails fatally after 10 attempts — unlike the OpAMP client, which has a schemeFlip retry for exactly this case. Confirmed via edge/internal/utils/enroll.go and edge/internal/supervisor/supervisor.go that no other consumer of this flag is affected.

How the data_dir issue was found

Live end-to-end testing on a fresh Linux ARM64 VM: agent enrolled successfully, a real (non-empty) config was delivered by edge-manager, but the worker never started — error while validating Observo Edge Worker config err: exit status 78, stderr: (empty stderr in the log, since it went to stdout). Running the worker's own validate command manually on the box surfaced the real message: data_dir "/opt/observo/data" does not exist. Creating the directory unblocked it immediately — worker validated cleanly and started, confirmed via process + healthy logs + real file-tailing activity.

Test plan

• bash -n install.sh / bash -n install_darwin.sh — syntax OK
• Verified live on VM: manually created /opt/observo/data, reran the worker's validate command → Validated, exit 0
• Restarted the agent service on the VM → worker started and stayed healthy
• Reviewer: confirm install.ps1 changes look right (not live-tested on Windows this session)

🤖 Generated with Claude Code