Linearization (.lin) files are written with tabular OutFmt (default ES10.3E2, ~4 significant digits), corrupting wide-dynamic-range Jacobians on disk

[mayankchetan]

Bug description

The glue code writes everything in linearization (.lin) files — the Jacobian matrices
(A/B/C/D, dUdu/dUdy, module blocks), the operating-point column, and the header scalars
(including Rotor Speed, which MBC3 postprocessors consume) — using the OpenFAST input's
tabular OutFmt, whose documented default "ES10.3E2" carries only ~4 significant digits.

For models whose state matrices span a wide dynamic range — e.g. BeamDyn blades with fully
populated sectional 6×6 mass/stiffness matrices, where the assembled A spans ~15 orders of
magnitude — 4 digits is not enough to round-trip the matrix through disk. Every downstream
eigenanalysis (openfast_toolbox / pyFAST, MBC3 / Campbell workflows, ACDC) then reports
physically impossible results for a parked, passive structure: spurious positive eigenvalues,
negative damping ratios, and merged or missing modes. The corruption is invisible in the
.lin file itself and is easily misread as a physics or solver defect (we initially did).

Evidence that this is purely a write-precision artifact:

1. Two decks differing only in OutFmt (ES10.3E2 vs ES18.9E3) produce clean vs
   corrupt eigenanalyses from the same simulation.
2. Re-truncating the wide-format matrices to 4 significant digits in numpy reproduces the
   corrupt eigenvalues to the digit.
3. A patched writer (fixed wide format) on a default-OutFmt deck is clean.

To Reproduce

1. Compile dev (double precision).
2. Linearize a parked BeamDyn-blade case with fully populated sectional matrices (e.g. an
   IEA reference blade) using the default OutFmt.
3. Run any MBC3/eigenanalysis on the .lin file.
4. Observe positive real parts / negative damping for a passive structure; repeat with
   OutFmt "ES18.9E3" and observe a clean spectrum.

Expected behavior

.lin numeric output round-trips the computed matrices at full working precision, independent
of the time-series formatting choice in OutFmt.

Fix

PR #3393 (f/lin-full-precision): all numeric .lin output uses a fixed
ES24.15E3 format; tabular time-series output continues to honor the user's OutFmt.
Workaround on released versions: set OutFmt "ES18.9E3" in any input file used for
linearization.

OpenFAST Version

dev @ 0160d29a0; double precision, gfortran, RelWithDebInfo, macOS (arm64). The defect is
long-standing (format usage predates current dev).

Additional context

Isolated during the BeamDyn rotating-frame linearization investigation (see companion issue #3394 on
spurious spin-softening with RotStates = True); the two defects are independent but both
corrupt BD Campbell diagrams, and decks used to study one must neutralize the other.

Developed with Anthropic Claude (Claude Code) under human direction.