fix: query default user missing group files

[AhmedBoutaraa57]

Issue #64: Query Default User Missing Group Files

Problem

When querying as the default user (username=None), data stored in the group path (tmp/group/{username}/...) is not accessible, leading to incomplete query results. This occurs because resolve_table_dir() only checks one path based on the username parameter.

Current Code Behavior

• Default user queries: Only check tmp/data/{db}/{table}/ (never checks group path)
• Group user queries: Only check tmp/group/{username}/{db}/{table}/ (never checks default path)
• cloud_fetch_parquet(): Always writes to group path (line 288)
• cloud_sink_parquet(): Merges local (default path) with cloud, but only reads from default path (line 242: build_files_list(..., None))
• cloud_sync_parquet(): Uploads from default path to cloud, downloads from cloud to group path, but never updates the default path with merged data

Data Flow Diagram

┌─────────────────────────────────────────────────────────────────────────┐
│                        DATA FLOW & PATH SEPARATION                       │
└─────────────────────────────────────────────────────────────────────────┘

                    ┌─────────────────────┐
                    │   Cloud Storage     │
                    │   (S3)              │
                    │   [A, B, C, D, E]   │
                    └──────────┬──────────┘
                               │
                               │ cloud_fetch_parquet(username="ahmed_test")
                               │ Line 288: Always writes to group path
                               ▼
        ┌──────────────────────────────────────────────┐
        │  tmp/group/ahmed_test/{db}/{table}/           │
        │  Files: activitydetails_2025-02-10.parquet   │
        │  Contains: Records A, B, C, D, E            │
        │  ✅ Accessible by: Group user queries        │
        │  ❌ NOT accessible by: Default user queries  │
        └──────────────────────────────────────────────┘
                               │
                               │ insert() writes new data
                               │ Line 436: Always writes to default path
                               ▼
        ┌──────────────────────────────────────────────┐
        │  tmp/data/{db}/{table}/                      │
        │  partition_date=2025-02-10/data.parquet      │
        │  Contains: Records F, G                      │
        │  ✅ Accessible by: Default user queries      │
        │  ❌ NOT accessible by: Group user queries    │
        └──────────────────────────────────────────────┘
                               │
                               │ cloud_sink_parquet()
                               │ Line 242: build_files_list(..., None)
                               │         → Only reads from tmp/data/...
                               │ Line 388-400: Merges local [F, G] with cloud [A-E]
                               │
                    ┌──────────┴──────────┐
                    │                     │
        ┌───────────▼──────────┐ ┌───────▼────────────┐
        │ Cloud Storage (S3)   │ │ Query Resolution  │
        │ After cloud_sink:    │ │ (resolve_table_dir)│
        │ [A, B, C, D, E, F, G]│ │                   │
        │ ✅ Complete (merged) │ │                   │
        └──────────────────────┘ │                   │
                                 │                   │
                    ┌────────────┴──────────┐       │
                    │                       │       │
        ┌───────────▼──────────┐ ┌───────────▼───────▼┐
        │ Default User Query  │ │ Group User Query  │
        │ username=None       │ │ username="user"   │
        │                     │ │                   │
        │ Returns: tmp/data/ │ │ Returns:          │
        │ Sees: F, G          │ │ tmp/group/...     │
        │ ❌ Missing: A-E      │ │ Sees: A, B, C, D, E│
        │                     │ │ ❌ Missing: F, G   │
        └─────────────────────┘ └───────────────────┘

Key Issues:
- cloud_sink_parquet() merges default path with cloud ✅
- BUT: cloud_sink_parquet() ignores group path data ❌
- Query resolution is EXCLUSIVE (only one path checked)
- Local paths remain split even though cloud is complete

Summary: Complete vs Incomplete Data

✅ Complete Data - Default User:
   - All inserts go to tmp/data/... (default path)
   - No data in tmp/group/...
   - Query sees all data ✅

✅ Complete Data - Group User:
   - cloud_fetch_parquet() downloads to tmp/group/...
   - No new inserts yet (or all inserts synced)
   - Query sees all data ✅

❌ Incomplete Data - Default User:
   - cloud_fetch_parquet() downloads to tmp/group/... [A, B, C]
   - User inserts new data → goes to tmp/data/... [D, E]
   - Query as default user → Only sees [D, E], missing [A, B, C] ❌
   
   OR
   
   - cloud_sync_parquet() downloads complete data to tmp/group/...
   - tmp/data/ still has old data (not updated)
   - Query as default user → Sees old data, missing new data from group path ❌

❌ Incomplete Data - Group User:
   - cloud_sync_parquet() downloads complete data to tmp/group/... [A, B, C]
   - User inserts new data → goes to tmp/data/... [D, E]
   - Query as group user → Only sees [A, B, C], missing [D, E] ❌
   - New inserts missed until next cloud_sync_parquet()

Root Cause:
- insert() always writes to tmp/data/... (default path)
- cloud_fetch_parquet() always writes to tmp/group/...
- Query checks only ONE path (default OR group, never both)
- cloud_sync_parquet() updates group path but NOT default path