Progress indicator during CSV loading

[vmvarela]

Description

When piping large CSV files, sql-pipe appears frozen — there is no feedback while rows are being inserted. A lightweight progress indicator on stderr eliminates the "is it working?" uncertainty.

Example

Loading... 125,000 rows\r
Loading... 250,000 rows\r
Loaded 312,450 rows in 1.2s

Acceptance Criteria

• During CSV loading, print a running row count to stderr using carriage return (\r) to update in place — no newlines until done
• Only shown when stderr is a TTY (suppressed in pipelines and scripts)
• Update frequency: every 10,000 rows (or configurable via constant) — not every row
• After loading completes, overwrite with final Loaded <n> rows in <t>s and a newline
• If --max-rows is set, show it as Loading... 5,000 / 100,000 rows
• Zero performance overhead when stderr is not a TTY
• README.md updated

Notes

• Depends on / pairs well with #U1 (row count after load) — can be done together
• Use std.os.isatty or equivalent in Zig to detect TTY
• Elapsed time via std.time.milliTimestamp()

Dependencies

Depends on #83 (row count must be tracked before a progress indicator can show it)