Decide: should a comment-only edit inside a grounded function count as GROUNDING_DRIFT?

[theyashasvipandey]

Question: should a comment-only edit inside a grounded function count as grounding drift?

Today it does. On Hono, adding one comment line inside errorHandler and then refreshing the
graph produced:

⚠ GROUNDING_DRIFT Grounded node body changed: …

GROUNDING_DRIFT compares the committed bodyHash with the node's current bodyHash. The body
hash covers the raw source text, so comments and formatting count.

For keeping it: a comment can be the thing the knowledge paraphrases, so a change there can
matter. It's also simple and has no false negatives.

Against: comment and formatting churn is common, especially from formatters and doc passes.
Each one produces a warning that can only be cleared by a sync session (see #124), which trains
people to ignore GROUNDING_DRIFT.

Options

1. Keep today's behaviour.
2. Hash a normalized body (comments and insignificant whitespace stripped, using the same
   tokenization the fingerprint already does) for bodyHash. That's a format change to a
   committed field, so it needs versioning like any bodyHash change.
3. Keep the raw hash, but also store a normalized hash. When only the raw one differs, report
   at a lower severity (info), or auto-accept.

Related: #124 (clearing GROUNDING_DRIFT needs an interactive AI session).