feat(cli): add sync observability with per-table progress and timing

[ealexisaraujo]

Summary

Improve nao sync observability so users always know what the CLI is doing during long-running database syncs, and pre-filter schemas to avoid unnecessary Snowflake queries.

Changes in provider.py

• Per-table progress — the progress bar description now shows SCHEMA → TABLE_NAME so users always see which table is being synced
• MofNCompleteColumn — progress bar shows 329/416 (completed/total) instead of just a percentage
• TimeElapsedColumn — elapsed time ticks in real-time as a built-in heartbeat, so users know the sync is alive even during slow queries
• Connection timing — logs how long db_config.connect() takes (e.g., Connected to my-db (1.9s))
• Schema discovery timing — logs schema count and duration (e.g., Found 3 schemas (297ms))
• Per-schema entry log — when entering a schema, shows matched vs total table count and list_tables() duration (e.g., ▸ TRANSFORM — 416 tables (of 416 total, listed in 1.2s))
• Slow query warnings — any accessor (columns/description/preview) taking >5 seconds is logged with its name and duration (e.g., ⏱ TRANSFORM.LARGE_TABLE description took 1m16s)
• Error context — errors now show which accessor failed, how long it ran before failing, and the error message (e.g., ✗ TRANSFORM.BAD_TABLE preview failed after 3.5s: Numeric value...)
• Per-schema completion summary — when a schema finishes, logs table count, duration, and error count (e.g., ✓ TRANSFORM — 416 tables synced in 38m0s (1 errors))
• Total error summary — if any errors occurred, shows the total count at the end
• Total sync duration — final summary includes overall time (e.g., 654 tables across 3 datasets in 50m11s)

Changes in snowflake.py

• Schema pre-filtering — new _schema_matches() method extracts schema prefixes from include/exclude patterns (e.g., ANALYTICS from ANALYTICS.*) and filters schemas before entering the per-table sync loop
• This avoids calling list_tables() to Snowflake for schemas that will have zero matching tables, reducing unnecessary API calls and warehouse cost

Motivation

When running nao sync on large databases (e.g., Snowflake with 100B+ row tables), several problems make the CLI difficult to use:

1. Progress bar appears stuck — it only updates per-schema, so within a schema with 400+ tables, users see no movement for 30+ minutes
2. No heartbeat — impossible to tell if the CLI is working on a slow COUNT(*) or has crashed
3. No timing data — users can't identify which tables/accessors cause slowdowns to make informed decisions about their config
4. Generic errors — when a template fails, the error doesn't say which accessor (columns/description/preview) failed or how long it ran
5. Wasted queries — get_schemas() returns all schemas from the database even when include patterns limit to a subset, causing unnecessary list_tables() calls for non-matching schemas

Why these accessors are slow

The three default templates have very different performance characteristics:

Accessor	Snowflake Query	Typical Duration
columns	INFORMATION_SCHEMA.COLUMNS metadata	Sub-second
description	SELECT COUNT(*) FROM table	Minutes on 100B+ row tables
preview	SELECT * FROM table LIMIT 10	Seconds, but can fail on type casting issues

Without timing, users had no visibility into which accessor was causing the sync to appear hung.

Example output (before vs after)

Before

🗄️  Syncing Databases
snowflake-prod: columns, description, preview

✗ Error generating preview.md for TRANSFORM.TABLE_WITH_BAD_DATA: Numeric value '11,380' is not recognized
⠦ snowflake-prod ━━━━━━━━━━━━━━━━━━━━━━━━━━━━╺━  94%
    SALES        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100%
⠦   TRANSFORM   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   1%

(No visibility into which table, which accessor, how long, or whether it's still alive)

After

🗄️  Syncing Databases
my-snowflake-db: columns, description, preview

  Connected to my-snowflake-db (1.9s)
  Found 3 schemas (297ms)
  ▸ ANALYTICS — 110 tables (of 110 total, listed in 621ms)
    ⏱ ANALYTICS.DIM_LARGE_TABLE description took 6.0s
  ✓ ANALYTICS — 110 tables synced in 4m57s
  ▸ SALES — 128 tables (of 128 total, listed in 597ms)
    ⏱ SALES.FACT_LARGE_TABLE description took 1m51s
  ✓ SALES — 128 tables synced in 7m14s
  ▸ TRANSFORM — 416 tables (of 416 total, listed in 1.2s)
⠋     TRANSFORM → STG_DIM_SOME_TABLE ━━━━━━━━━╸━━━━━ 329/416 79% 0:30:31
    ✗ TRANSFORM.BAD_TABLE preview failed after 3.5s: Numeric value '11,380' is not recognized
    ⏱ TRANSFORM.VERY_WIDE_TABLE description took 1m16s
    ⏱ TRANSFORM.VERY_WIDE_TABLE preview took 3m42s
  ✓ TRANSFORM — 416 tables synced in 38m0s (1 errors)
  ⚠ 1 total errors during sync

Files changed

File	Change
cli/nao_core/commands/sync/providers/databases/provider.py	Add timing, per-table progress, MofNCompleteColumn, TimeElapsedColumn, slow query warnings, error context, schema/total summaries
cli/nao_core/config/databases/snowflake.py	Add _schema_matches() for schema-level pre-filtering based on include/exclude patterns

Test plan

• Run nao sync -p databases against a Snowflake database with multiple schemas and hundreds of tables
• Verify connection timing appears on startup
• Verify schema count reflects only included schemas (not all database schemas)
• Verify per-table name shows in progress bar during sync
• Verify MofNCompleteColumn shows N/M count (e.g., 329/416)
• Verify elapsed time column ticks in real-time
• Verify slow queries (>5s) produce a warning with accessor name and duration
• Verify errors show accessor name, duration, and error message
• Verify per-schema completion summary appears with table count, duration, and error count
• Verify total sync duration appears in final summary
• Verify make lint passes (ty, ruff check, ruff format)
• Verify empty include/exclude lists still return all schemas (no regression)

Fixes #206

🤖 Generated with Claude Code

[MatLBS]

Hi ealexisaraujo 👋, thank you so much for your PR,

I tried it and everything works well