Remove insert-vs-alter transaction conflict check

Concurrent INSERT and ALTER TABLE operations (e.g., ADD COLUMN, RENAME
COLUMN) on the same table should not conflict. Each data file carries
its own mapping_id that describes the column layout at write time. The
multi-file reader uses this mapping to correctly read files written
under older schemas, so an INSERT that was planned before an ALTER
can safely commit after it.

This is needed for tools like sqlmesh that perform schema evolution
(ALTER TABLE) on target tables while concurrent data pipelines INSERT
into those same tables. Previously, the INSERT would fail with
"Transaction conflict - attempting to insert into table - but another
transaction has altered it" after exhausting all retries.

The insert-vs-drop conflict check is preserved since a dropped table
is genuinely gone.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: fuziontech

src/storage/ducklake_transaction.cpp

@@ -1232,11 +1232,13 @@ void DuckLakeTransaction::CheckForConflicts(const TransactionChangeInformation &
 	}
 	for (auto &table_id : changes.tables_inserted_into) {
 		ConflictCheck(table_id, other_changes.dropped_tables, "insert into table", "dropped it");
-		ConflictCheck(table_id, other_changes.altered_tables, "insert into table", "altered it");
+		// NOTE: insert-vs-alter is safe because each data file carries its own mapping_id
+		// that describes the column layout at write time. The multi-file reader uses
+		// this mapping to correctly read files written under older schemas, even after
+		// ALTER TABLE ADD/DROP/RENAME COLUMN operations.
 	}
 	for (auto &table_id : changes.tables_inserted_inlined) {
 		ConflictCheck(table_id, other_changes.dropped_tables, "insert into table", "dropped it");
-		ConflictCheck(table_id, other_changes.altered_tables, "insert into table", "altered it");
 	}
 	for (auto &table_id : changes.tables_deleted_from) {
 		ConflictCheck(table_id, other_changes.dropped_tables, "delete from table", "dropped it");

test/sql/transaction/transaction_insert_alter_no_conflict.test

@@ -0,0 +1,103 @@
+# name: test/sql/transaction/transaction_insert_alter_no_conflict.test
+# description: Test that concurrent INSERT and ALTER TABLE do not conflict
+# group: [transaction]
+
+require ducklake
+
+require parquet
+
+test-env DUCKLAKE_CONNECTION __TEST_DIR__/{UUID}.db
+
+test-env DATA_PATH __TEST_DIR__
+
+statement ok
+ATTACH 'ducklake:${DUCKLAKE_CONNECTION}' AS ducklake (DATA_PATH '${DATA_PATH}/ducklake_insert_alter_no_conflict_files')
+
+statement ok
+SET immediate_transaction_mode=true
+
+statement ok
+CREATE TABLE ducklake.test(i INTEGER, j INTEGER);
+
+statement ok
+INSERT INTO ducklake.test VALUES (1, 10);
+
+# one transaction inserts while another alters: should NOT conflict
+statement ok con1
+BEGIN
+
+statement ok con2
+BEGIN
+
+statement ok con1
+INSERT INTO ducklake.test VALUES (2, 20);
+
+statement ok con2
+ALTER TABLE ducklake.test ADD COLUMN k INTEGER
+
+statement ok con1
+COMMIT
+
+statement ok con2
+COMMIT
+
+# verify both changes applied: 2 rows, 3 columns, new column is NULL for existing rows
+query III
+SELECT i, j, k FROM ducklake.test ORDER BY i
+----
+1	10	NULL
+2	20	NULL
+
+# one transaction alters while another inserts (reversed order): should NOT conflict
+statement ok con1
+BEGIN
+
+statement ok con2
+BEGIN
+
+statement ok con1
+ALTER TABLE ducklake.test ADD COLUMN m INTEGER
+
+statement ok con2
+INSERT INTO ducklake.test VALUES (3, 30, 300);
+
+statement ok con1
+COMMIT
+
+statement ok con2
+COMMIT
+
+# verify: 3 rows, 4 columns
+query IIII
+SELECT i, j, k, m FROM ducklake.test ORDER BY i
+----
+1	10	NULL	NULL
+2	20	NULL	NULL
+3	30	300	NULL
+
+# concurrent insert and rename column: should NOT conflict
+statement ok con1
+BEGIN
+
+statement ok con2
+BEGIN
+
+statement ok con1
+INSERT INTO ducklake.test VALUES (4, 40, 400, 4000);
+
+statement ok con2
+ALTER TABLE ducklake.test RENAME COLUMN m TO n
+
+statement ok con1
+COMMIT
+
+statement ok con2
+COMMIT
+
+query IIII
+SELECT i, j, k, n FROM ducklake.test ORDER BY i
+----
+1	10	NULL	NULL
+2	20	NULL	NULL
+3	30	300	NULL
+4	40	400	4000