fix: PMT crashes in a single-channel Simulation (float() on a (1, 1, 1, 1) exposure)

[federico-carrara]

Description

Simulation.digital_image() crashes with a PMT detector when the simulation has a single channel:

import numpy as np
import microsim.schema as ms
from microsim.schema.detectors import lib

truth = np.ones((1, 64, 64))
sim = ms.Simulation(
    truth_space=ms.ShapeScaleSpace(shape=truth.shape, scale=(0.1, 0.1, 0.1)),
    sample=ms.Sample(labels=[ms.FluorophoreDistribution.from_array(truth)]),
    modality=ms.Identity(),
    detector=lib.PMT_GAASP,  # default channels: [FITC]
    exposure_ms=0.01,
)
sim.digital_image()
# TypeError: only 0-dimensional arrays can be converted to Python scalars

The exposure time is a single number, but by the time it reaches the PMT it is a 4-d array.

Call chain:

1. Simulation.digital_image(): called without exposure_ms, it builds one exposure per channel as a DataArray over c (shape (1,)) and passes it to detector.render() → _Detector.simulate().
2. _Detector.simulate(): exposure_s = exposure_ms / 1000 * n_avg (still (1,)), then _per_channel(exposure_s, ndim) from _detector.py adds trailing axes so that it broadcasts against the (C, Z, Y, X) image: shape (1, 1, 1, 1). This goes to _after_multiplication() as time_s.
3. PMT._after_multiplication(): passes time_s (saturation, if max_output_current_ua is set) or dwell_s (amplifier bandwidth, if time_constant_us > 0) as dt to the _per_channel in _point.py. That is a different helper with the same name.
4. _per_channel() in _point.py: dt.size == 1 is true, so it calls float(dt) on the (1, 1, 1, 1) array. float() on an array with ndim > 0 has been deprecated since NumPy 1.25, and it raises in NumPy 2.4 (tested with 2.4.6).

Why the tests didn't catch it: the crash needs all four of these at once:

• exposures passed per channel. A float exposure_ms is never reshaped in step 2, and every test in test_point_detectors.py calls render() with a float.
• exactly one channel. With C > 1 the else branch runs, and it flattens dt with dt.ravel().
• a PMT with saturation or a bandwidth limit. Both PMT presets set max_output_current_ua. HyD and the cameras don't use this helper.
• a NumPy version that raises. Older versions only emit a DeprecationWarning.

Changes

• _per_channel() in _point.py: float(dt) → float(dt.item()). .item() returns the single value whatever the number of axes.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.15%. Comparing base (0aea364) to head (f176506).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #159   +/-   ##
=======================================
  Coverage   88.15%   88.15%           
=======================================
  Files          48       48           
  Lines        3292     3292           
=======================================
  Hits         2902     2902           
  Misses        390      390           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tlambert03]

Good catch thanks!