fix(core): report output_dim, not min_output_dim, when it is too small

[ChrisW09]

Summary

With adaptive=False, setting an output_dim below the family floor produces an error that
blames min_output_dim — a parameter the user never set, and one that is ignored in
non-adaptive mode.

Reproduction

import numpy as np, pandas as pd
from pretab import Preprocessor

df = pd.DataFrame({"a": np.random.default_rng(0).normal(size=50)})
y = np.random.default_rng(1).normal(size=50)

for method in ["ple", "cubicspline"]:
 try:
 Preprocessor(numerical_method=method, output_dim=0).fit(df, y)
 except Exception as e:
 print(f"{method:14} -> {str(e).splitlines()[0]}")

ple -> min_output_dim must be >= 1, got 0.
cubicspline -> output_dim must be >= 3 for the cubic spline basis, got 0

Both were given output_dim=0. The families that pre-validate output_dim themselves report
it correctly; those that rely on the shared bounds resolver do not.

The follow-up line compounds it: "Fix: raise min_output_dim to at least the family minimum"
— acting on that advice would not help, since min_output_dim is unused when
adaptive=False.

Expected

The message names the parameter the caller actually set.

Actual

min_output_dim is named whenever the floor check trips, regardless of which parameter
produced the offending value.

Root cause

pretab/core/adaptive.py:89-94:

label = floor_label if floor_label is not None else str(floor)
if lo < floor:
 raise InvalidParamError(
 f"min_output_dim must be >= {label}, got {lo}.\n"
 "Fix: raise min_output_dim to at least the family minimum."
 )

lo is set a few lines earlier to output_dim on the non-adaptive branch
(adaptive.py:84), so the message is only correct when adaptive=True and the user
supplied min_output_dim.

Suggested fix

Pick the name from the branch that produced lo:

if lo < floor:
 name = "min_output_dim" if self.adaptive and min_req is not None else "output_dim"
 raise InvalidParamError(
 f"{name} must be >= {label}, got {lo}.\n"
 f"Fix: raise {name} to at least the family minimum."
 )

Cosmetic, but it is the first thing a user sees when they mis-size a transformer, and it
currently points them at the wrong knob.

Environment

• pretab 0.1.0 (main @ 51c3043)
• Python 3.11.15, numpy 2.4.6, pandas 2.3.3, scikit-learn 1.9.0, scipy 1.17.1
• macOS (darwin 25.5.0)

[mkumar73]

fixed in v1.0.0