Warn on duplicate column names

[vmvarela]

Description

When a CSV has duplicate column names (e.g., two columns both named value), sql-pipe silently renames them by appending a suffix (value, value_1, value_2). The user has no way to know this happened and may write queries against the original names that silently select the wrong column.

Example

$ printf 'id,value,value\n1,a,b' | sql-pipe 'SELECT value_1 FROM t'
b

The user expected column value but had to guess the renamed form value_1.

Acceptance Criteria

• When one or more column names are deduplicated, print a warning to stderr listing each rename: warning: duplicate column "value" renamed to "value_1"
• Warning is written to stderr so it does not corrupt stdout (CSV/JSON output)
• Warning is suppressed when stderr is not a TTY (pipe mode) — or printed always, consistently; decide and document
• Existing tests still pass; add a test for the warning message

Notes

• Deduplication logic is in src/main.zig around lines 356–364
• Format: warning: duplicate column "<original>" renamed to "<new>" (one line per rename)