Persist tmp/opengeometadata across deploys with Capistrano linked_dirs

[srappel]

Summary

opendataharvest currently expects to read/write harvested OGM data under tmp/opengeometadata, but in production that directory lives inside the timestamped release unless it is explicitly linked into shared/.

This creates an awkward split between the app-relative config path and the production path we actually want to persist across deploys.

Problem

Right now:

• config/opendataharvest.yaml uses tmp/opengeometadata
• production behavior appears to want /var/www/rubyapps/uwm-geoblacklight/shared/tmp/opengeometadata
• tmp/opengeometadata is not currently included in Capistrano linked_dirs

That means harvest data may end up release-local unless the config is manually changed on the server, which introduces config drift and makes the pull/index workflow more fragile.

Proposed direction

Treat tmp/opengeometadata like the other release-unsafe directories already managed by Capistrano:

• keep the app config path as tmp/opengeometadata
• add tmp/opengeometadata to Capistrano linked_dirs
• let production resolve that path to shared/tmp/opengeometadata via symlink

Why this matters

• preserves harvested OpenGeoMetadata data across releases
• avoids hardcoding a production-only absolute path in app config
• reduces config drift between environments
• makes full pull/index tasks operate on a stable shared path in production