fix(clickhouse): encode schema names in JDBC URLs

[HandSonic]

Summary

• Encode the selected ClickHouse schema as one UTF-8 JDBC URL path segment before reconnecting.
• Prevent /, ?, #, spaces, Unicode, plus signs, and literal percent text from changing URL structure or the selected schema.
• Preserve the existing JDBC query suffix and add focused regression coverage.

Validation

• Focused regression tests: 3 passed.
• Full ClickHouse module tests: 46 passed, 0 failed.
• ClickHouse reactor package: passed.
• Fork code CI: frontend, backend, repository checks, Java/JavaScript CodeQL, SBOM, and license checks passed.
• Independent adversarial review: no findings.

Follow-up regression fix related to #2685 for schema names containing reserved URL characters. This intentionally does not close that issue; the separate IPv6 URL fix is in #2690.

Latest-main verification (2026-09-04)

• Rebased onto 144a04ee2; full ClickHouse plugin suite passed (55 tests), including IPv6, user-info, query/fragment, schema encoding, and idempotence cases.n- git diff --check` and pairwise integration analysis passed.

[openai0229]

Blocking integration issue: this branch still applies encoding on top of the old implementation. That implementation is exactly what #2690 replaces because it cannot reliably preserve IPv6 authorities, no-path URLs, fragments, and other URL shapes. Merging the two fixes independently leaves overlapping implementations and an avoidable conflict.\n\nPlease base this change on #2690's helper and encode the schema as one UTF-8 path segment inside that helper. Keep the reserved-character tests from this PR and the IPv6/query/fragment tests from #2690 so there is one URL-rewrite implementation.

Superseded by a corrected review because shell quoting corrupted the submitted text.

[openai0229]

Blocking integration issue: this branch still applies encoding on top of the old url.split(connectAddress) implementation. That implementation is exactly what #2690 replaces because it cannot reliably preserve IPv6 authorities, no-path URLs, fragments, and other URL shapes. Merging the two fixes independently leaves overlapping implementations and an avoidable conflict.

Please base this change on #2690's replaceDatabaseInJdbcUrl helper and encode the schema as one UTF-8 path segment inside that helper. Keep the reserved-character tests from this PR and the IPv6/query/fragment tests from #2690 so there is one URL-rewrite implementation.

[HandSonic]

Addressed on the current head: there is now one
eplaceDatabaseInJdbcUrl implementation with both the schema-as-one-UTF-8-segment cases and the IPv6/no-path/query/fragment/idempotence cases. The strict-subset #2690 was closed. The full ClickHouse suite passes (55 tests) on current main. Please re-review when convenient.