feat: add diagnostics for EllipseCloud.rescale()

[t-uda]

Motivation

Users currently get only the returned scalar from EllipseCloud.rescale() and must compute their own diagnostics to understand what the scale factor means. This slows down tuning in applications where ellipsoid size must be compared against spatial point spacing.

Scope

• Design a lightweight diagnostics interface for rescale().
• Prefer a non-verbose, programmatic result over print-only output.

Acceptance Criteria

• Users can inspect the scale factor together with the pre/post summary values used to justify it.
• The default return path stays backward compatible.
• Documentation includes a short example showing how to interpret the diagnostics.

Open Design Question

Decide whether diagnostics should be exposed via a new method, a small dataclass return value, or an opt-in return_diagnostics=True flag.

[t-uda]

Triage note (2026-07): the implementation for this issue exists in PR #119 (frozen RescaleDiagnostics dataclass + opt-in return_diagnostics=False flag — the flag option from the open design question). CI is green and the branch targets the release branch ellphi-0.1.3. This issue will be closed manually when #119 merges (no auto-close link exists). The duplicate PR #120 has been closed.

Interaction to keep in mind: #128 generalises rescale() to arbitrary dimensions — after #119 merges, the diagnostics summaries must be made dimension-generic as part of #128 (noted there).

[t-uda]

Resolved by PR #119 (merged into ellphi-0.1.3): EllipseCloud.rescale() now supports return_diagnostics=True, returning a frozen RescaleDiagnostics dataclass (scale + pre/post summaries), with .pyi overloads and tests. Default return is unchanged (backward compatible).

Follow-up already tracked: #128 must make the diagnostics dimension-generic when generalising rescale() to arbitrary dimensions.