Add SQLFetchScroll support for fetchall() and fetchmany() - Fixes #1012

[bchoudhary6415]

Added rowset-based fetching using SQLFetchScroll(SQL_FETCH_NEXT) for ibm_db.fetchall() and ibm_db.fetchmany() APIs. This replaces the previous row-by-row SQLFetch() approach, improving performance by fetching multiple rows per ODBC/CLI call using array-bound column buffers.

Fixes #1012

---

What's New

APIs to support SQLFetchScroll

• ibm_db.fetchall(stmt) — Fetches all remaining rows from the result set as a list of tuples.
• ibm_db.fetchmany(stmt, size) — Fetches up to size rows from the result set as a list of tuples.

Both APIs also work through the DBI layer: cursor.fetchall() and cursor.fetchmany(size).

SQL_ATTR_ROW_ARRAY_SIZE Support

A new statement attribute SQL_ATTR_ROW_ARRAY_SIZE controls how many rows are fetched per SQLFetchScroll call (i.e., the rowset size).

Feature	Details
Default value	1000
Minimum	1
Maximum	Automatically capped if total buffer exceeds 64 MB
Set via	ibm_db.set_option(stmt, {ibm_db.SQL_ATTR_ROW_ARRAY_SIZE: N}, 0)
Get via	ibm_db.get_option(stmt, ibm_db.SQL_ATTR_ROW_ARRAY_SIZE, 0)

How it works:

• SQL_ATTR_ROW_ARRAY_SIZE tells the ODBC driver how many rows to fetch in a single SQLFetchScroll call.
• The driver binds column buffers large enough to hold N rows, then fetches N rows at once into those buffers.
• Fewer round-trips to the CLI layer = better performance, especially for large result sets.
• If not set by the user, defaults to 1000 rows per fetch call.

Example:

import ibm_db

conn = ibm_db.connect(conn_str, "", "")
stmt = ibm_db.exec_immediate(conn, "SELECT * FROM LARGE_TABLE")

# Use default rowset size (1000)
rows = ibm_db.fetchall(stmt)

# Or set custom rowset size for fine-tuning
stmt2 = ibm_db.exec_immediate(conn, "SELECT * FROM LARGE_TABLE")
ibm_db.set_option(stmt2, {ibm_db.SQL_ATTR_ROW_ARRAY_SIZE: 5000}, 0)
rows = ibm_db.fetchall(stmt2)

# fetchmany - fetch in batches
stmt3 = ibm_db.exec_immediate(conn, "SELECT * FROM LARGE_TABLE")
batch = ibm_db.fetchmany(stmt3, 100)  # fetch 100 rows  

ibmdb_tests.py Fix

Fixed ValueError: Empty module name that occurred when running test files with SINGLE_PYTHON_TEST environment variable set to blank/empty. The load_tests() function now filters out empty module names before attempting import:
tests = [ t for t in tests if t ] # Filter out any empty module names

Note: LOB Column Handling

When any column in the result set is a LOB type (BLOB, CLOB, DBCLOB, XML), the fetch automatically falls back to single-row SQLFetch() + SQLGetData(). This is because:
LOB columns cannot be array-bound with SQLBindCol()
SQLGetData() is required for LOB data retrieval and only works on the current row
This fallback is transparent to the user — fetchall() and fetchmany() return the same results regardless of whether the table contains LOB columns.