SQLite driver returns an incorrect row count (-1) for UPDATE statements that begin with a comment.

Downstream Reference:
https://github.com/sqlalchemy/sqlalchemy/issues/4396


Test Case:

```
import sqlite3

conn = sqlite3.connect(":memory:")

cursor = conn.cursor()

cursor.execute("""
CREATE TABLE foo (
    id INTEGER NOT NULL,
    updated_at DATETIME,
    PRIMARY KEY (id)
)
""")

cursor.execute("""
/* watermarking bug */
INSERT INTO foo (id, updated_at) VALUES (?, ?)
""", [1, None])


cursor.execute("""
    UPDATE foo SET updated_at=? WHERE foo.id = ?
""", ('2018-12-02 14:55:57.169785', 1))

assert cursor.rowcount == 1

cursor.execute("""
    /* watermarking bug */
    UPDATE foo SET updated_at=? WHERE foo.id = ?
""", ('2018-12-03 14:55:57.169785', 1))

assert cursor.rowcount == 1
```

Did some debugging here. If I am understanding this correctly the rowcount is set at https://github.com/python/cpython/blob/b8e689a6e8134e88f857a55e50b6a4977967e385/Modules/_sqlite/cursor.c#L574 . It checks for is_dml flag that is set here https://github.com/python/cpython/blob/b8e689a6e8134e88f857a55e50b6a4977967e385/Modules/_sqlite/statement.c#L78

The part to set is_dml skips space, tabs and newline and checks for the first set of characters that is not skipped to be insert, update, delete or replace and in this case the first set of characters to be matched will be "/* watermarking */". Thus with comment not matching, is_dml is not set and -1 is set for the rowcount.

I took a stab at patch. It fixes the issue for me, as proven via the Test Case.

See also bpo-36859.