Fix issue duckdb#14659 (duckdb#15411)

Fix: duckdb#14659

Author: hannes

Diff for: data/csv/repromarket.csv

@@ -0,0 +1,3 @@
+[email protected]:krlleta
+[email protected]:bogus:vega
+[email protected]:mirko

Diff for: src/execution/operator/csv_scanner/sniffer/dialect_detection.cpp

@@ -234,7 +234,7 @@ void CSVSniffer::AnalyzeDialectCandidate(unique_ptr<ColumnCountScanner> scanner,
 	idx_t num_cols = sniffed_column_counts.result_position == 0 ? 1 : sniffed_column_counts[0].number_of_columns;
 	const bool ignore_errors = options.ignore_errors.GetValue();
 	// If we are ignoring errors and not null_padding , we pick the most frequent number of columns as the right one
-	bool use_most_frequent_columns = ignore_errors && !options.null_padding;
+	const bool use_most_frequent_columns = ignore_errors && !options.null_padding;
 	if (use_most_frequent_columns) {
 		num_cols = sniffed_column_counts.GetMostFrequentColumnCount();
 	}
@@ -355,7 +355,7 @@ void CSVSniffer::AnalyzeDialectCandidate(unique_ptr<ColumnCountScanner> scanner,
 	// - There's a single column before.
 	// - There are more values and no additional padding is required.
 	// - There's more than one column and less padding is required.
-	if (columns_match_set && rows_consistent &&
+	if (columns_match_set && (rows_consistent || (set_columns.IsSet() && ignore_errors)) &&
 	    (single_column_before || ((more_values || more_columns) && !require_more_padding) ||
 	     (more_than_one_column && require_less_padding) || quoted) &&
 	    !invalid_padding && comments_are_acceptable) {

Diff for: test/sql/copy/csv/auto/test_csv_auto.test

@@ -8,6 +8,24 @@ PRAGMA enable_verification
 statement ok 
 PRAGMA verify_parallelism
 
+query II
+FROM read_csv('data/csv/repromarket.csv',
+   columns={
+      'email': 'varchar',
+      'password': 'varchar'
+   },
+   all_varchar=true,
+   delim=':',
+   header=false,
+   skip=0,
+   null_padding=true,
+   ignore_errors=true
+);
+----
+[email protected]	krlleta
+[email protected]	NULL
+[email protected]	mirko
+
 query I
 FROM read_csv('data/csv/pipe_delim.csv', columns={'a': 'VARCHAR'}, auto_detect=False)
 ----