physionet_challenge_2012: implausible values reach the features #148

Problem

PhysioNet warns that "occasional outliers should be expected", and source preparation averages them in. In the prepared source:

• Blood pressures of 0 mmHg: SysABP 50 stays, DiasABP 51, NISysABP 34, MAP 2, NIMAP 1, NIDiasABP 2.
• pH above 8 in 11 stays (up to 128.5) and 0 in one; Temp below 30 °C in 3 (down to 21.6); Weight below 20 kg in 6 (down to 1.78); Height below 100 cm or above 250 cm in 28 (up to 462).

Z-normalization then compresses every plausible value of these features.

Scope

• Source preparation and dataset generation for physionet_challenge_2012; both budget settings.
• Changes paper results: it is in the kdd26 dataset list.

Options to decide

• Treat values outside physiological ranges as missing in source preparation, before averaging (needs a documented range per variable), or in _read_source.
• Or keep them and document them (the dataset page lists examples today).

Acceptance criteria

• The handling of out-of-range values is decided, implemented, and stated on docs/reference/datasets/physionet_challenge_2012.md.
• just qa passes.