fix: include rotor inertia from spanwise blade mass in auto-RNA (#130)

CHANGELOG.md

@@ -8,7 +8,30 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ## [Unreleased]
 
-(nothing yet)
+### Fixed
+
+- **WindIO auto-RNA now carries the rotor inertia from the spanwise blade
+  mass (#130).** `read_windio_rna` (and `lumped_rna_cal=True`) previously
+  lumped the blades as a bare point mass at the rotor apex, which captured
+  the hub-to-tower-top translation but dropped the rotor's own diametral
+  inertia from the blade mass being spread along the span. That term is the
+  dominant part of the rotor's contribution to the tower-top rotary
+  inertia (on IEA-22 it is ~2.5x the value the point-mass lump produced),
+  so tower fore-aft / side-side frequencies from a rigid-RNA lump came out
+  too high versus a rigid-rotor reference. The rotor is now assembled as a
+  rigid body, `diag([I_polar, I_diam, I_diam])` about the rotor centre of
+  mass with `I_polar = N_bl · ∫ (dm/ds) · r² ds` and the in-plane lever
+  `r = (hub_radius + span)·cos(cone)` (the coned pitch-axis distance, with
+  prebend and sweep folded in), using the hub diameter and cone angle when
+  present. A coned rotor's mass sits off the hub plane, so the rotor is
+  placed at its true centre of mass before the parallel-axis shift; this
+  also moves the tower-top centre of mass slightly (by the precone term)
+  for a non-zero `cone_angle`. Only each blade's own sectional spin inertia
+  is still excluded. This changes the auto-RNA inertia (and hence coupled
+  tower frequencies), and the CM for coned rotors, relative to 1.16.0; the
+  total mass is unchanged. This intentionally goes beyond the ElastoDyn
+  deck path's point-mass lumping, which the WindIO per-station blade mass
+  makes possible.
 
 ## [1.16.0] — 2026-07-03