fix(cursor): make whatnext() consistent with next() at EOF

[Aggarwalmansi]

Summary

whatnext() was returning a cursor state that does not occur in actual execution when reaching the final iteration.
This caused a mismatch between predictive (whatnext) and actual (next) behavior, leading to incorrect EOF signaling—especially for partitioned runs.

What was failing

A unit test in runner-cursor.test.js exposed that at the final position:

• whatnext() predicted EOF
• while next() did not yet signal EOF

This divergence caused incorrect cursor state prediction.

What changed

• Preserve actual cursor position and iteration in whatnext()
• Adjust EOF signaling only when partition cycles are active
• No changes to next() or cursor mutation logic

Tests

• Existing unit test failed before this change
• All cursor-related tests now pass and predictive behavior matches runtime behavior