CSV / JSON export (and import) #3

Motivation: Dives live in a local SQLite file. Users need a way
to back up, share, and migrate. CSV and JSON are the lowest common
denominator for logbook portability.

What to build:

• File menu → Export → CSV / JSON (all dives, or filtered by current
  search)
• File menu → Import → CSV (creates dives from a file matching
  the export schema)
• CSV columns: dive_date, start_time, end_time, dive_time_minutes,
  max_depth_m, avg_depth_m, air_temp_c, water_temp_c, visibility_m,
  start_pressure_bar, end_pressure_bar, sites (semicolon-separated),
  buddies (semicolon-separated as name:role), notes
• JSON: same fields, nested arrays for sites / buddies
• Export dialog: "All dives" / "Filtered to current search" toggle
• Import dialog: file picker, preview first 5 rows, confirm
• Use the existing find_or_create semantics for sites and buddies
  on import (no duplicates by name+country or normalized name)

Acceptance criteria:

• Round-trip: export → wipe DB → import produces identical rows
  (modulo ids)
• CSV opens correctly in Excel / Numbers / Google Sheets
• JSON validates with json.loads
• Import shows progress for files with 100+ rows
• 3 new tests: export-empty, export-nonempty, import-roundtrip

Related: #17 (test data generator)