[docs] Clarify sequence-group behavior when combined with aggregate functions

[dyzcs]

Purpose

Linked issue: related to #5179

Users are frequently confused when sequence-group is combined with aggregate functions (e.g. sum, last_non_null_value), expecting it to act as a version filter (records with a smaller sequence value are ignored). However, when aggregate functions are involved, the sequence-group field acts as an ordering key: every record with a non-NULL sequence value participates in the aggregation, regardless of whether its value is larger or smaller than the currently stored one.

This PR clarifies the distinction by:

1. Adding an informational note at the top of the "Aggregation For Partial Update" section that explicitly explains the behavioral difference between using sequence-group with and without aggregate functions.
2. Fixing a misleading code comment -- a should not be updated in an existing example, which contradicted the shown output (a=6 via sum aggregation).

Tests

Documentation-only change. No tests required.

API and Format

No API or storage format changes.

Documentation

This PR itself is a documentation improvement. No additional documentation needed.

Generative AI tooling

Generated-by: Claude (Cursor AI assistant), used to help draft and review the documentation wording.