Fix(clone): allow port in DB_BASE_URL url

[wezell]

This fixes issue #35505 and properly splits if there is a port specified
in the url

This PR fixes: #35505

[claude]

Claude finished @wezell's task in 59s —— View job

---

PR Review

• Read changed file and diff
• Analyze URL parsing logic for edge cases
• Post review findings

---

The core fix is correct — the %%:* / ##*: / equality-check pattern for splitting host[:port] is sound and handles both the no-port and with-port cases properly. The -p "${DB_PORT}" is consistently added to all four psql calls.

Issues worth noting:

1. DB_BASE_URL is parsed before it's validated.
   Lines 149–157 run unconditionally at script start. The if [ -z $DB_BASE_URL ] guard is at line 172 — well after parsing. If DB_BASE_URL is empty/unset, DB_HOST, DB_PORT, and DB_NAME all end up empty or at defaults, and psql gets called with -d "" (confusing error). Not introduced here, but the new parsing block is a good place to move the guard earlier.

2. IPv6 hosts will break with the new parsing.
   For jdbc:postgresql://[::1]:5432/dbname, DB_HOSTPORT = [::1]:5432, and ${DB_HOSTPORT%%:*} strips from the first colon, giving DB_HOST="[". The old code (${DB_HOST%%/*}) was IPv6-safe. Probably not a real-world concern in Docker deployments, but worth a note.

3. DB_NAME can include query parameters (pre-existing).
   jdbc:postgresql://host/dbname?ssl=true → DB_NAME="dbname?ssl=true" → psql fails. Not introduced here.

4. Unquoted variable in validation (pre-existing).
   Line 172: if [ -z $DB_BASE_URL ] should be if [ -z "$DB_BASE_URL" ].

None of these are blockers for the stated fix. Items 2–4 are pre-existing; item 1 is a mild ordering concern. The fix does what it says.