fix(skill/databricks-genie): use jq for cross-workspace catalog remap

[jacksandom]

Summary

sed -i '' in the Cross-Workspace Migration section is BSD/macOS-only — the
empty-string argument that's required by BSD sed is rejected by GNU sed on
Linux. Linux users running the documented snippet hit sed: -e expression #1, char 0: no previous regular expression.

Replace with jq, which is portable across macOS and Linux, already used
throughout the skill, and operates on JSON values directly.

Diff

-When migrating between workspaces, catalog names often differ. Export the space, remap with `sed`, then import:
+When migrating between workspaces, catalog names often differ. Export the space, remap, then import. Use `jq` to rewrite identifiers — it's portable across macOS/Linux (unlike `sed -i`, whose in-place flag differs between BSD and GNU) and operates on JSON values directly:

-sed -i '' 's/source_catalog/target_catalog/g' genie_space.json
+jq --arg src source_catalog --arg dst target_catalog \
+   '(.. | strings) |= gsub($src; $dst)' genie_space.json \
+   > genie_space.tmp && mv genie_space.tmp genie_space.json

Evaluation

Ran stf audit (L3 — static SKILL.md quality audit) against both this branch
and origin/experimental via the Databricks AI Gateway:

Dimension	Experimental	This PR	Δ
tool_accuracy	10.0	10.0	0
examples_valid	10.0	10.0	0
self_contained	8.0	8.0	0
no_conflicts	9.0	9.0	0
security	8.0	8.0	0
llm_navigable_structure	9.0	9.0	0
actionable_instructions	9.0	9.0	0
scoped_clearly	8.0	8.0	0
error_handling	7.0	7.0	0
no_hallucination_triggers	7.0	7.0	0
Overall	0.85	0.85	0

NEUTRAL on L3 — the rubric doesn't grade for cross-platform shell portability,
so the actual fix isn't visible at this granularity. No regression introduced.
stf lint clean. No MCP refs in the skill, so L1 MCP-preflight is N/A.

How to verify

# Smoke-test the new jq snippet
cat > /tmp/genie_space.json <<'EOF'
{"data_sources":{"tables":[{"identifier":"source_catalog.ops.t1"}]}}
EOF
jq --arg src source_catalog --arg dst target_catalog \
   '(.. | strings) |= gsub($src; $dst)' /tmp/genie_space.json \
   > /tmp/out.json && cat /tmp/out.json
# → identifier rewritten to target_catalog.ops.t1

# Reproduce the L3 audit
stf audit databricks-skills/databricks-genie

Test plan

• stf lint databricks-skills/databricks-genie clean
• stf audit PASS on this branch (0.85)
• stf audit PASS on origin/experimental (0.85) — no regression
• jq snippet verified on a sample genie_space.json (rewrites both identifier fields and catalog names embedded in instruction text)