Pass down the same KLIP parameters in contrast analysis

[kglidic]

JWSTData, the pyKLIP instrument class used for JWST observations, accepts several parameters, including filepaths, psflib_filepaths, center_keywords, center_include_offset, and highpass. In run_obs(), the dataset is initialized as:

dataset = JWSTData(
    filepaths,
    psflib_filepaths,
    highpass=kwargs_temp['highpass'],
    center_include_offset=False,
    center_keywords=['STARCENX', 'STARCENY']
)

dataset.IWA = kwargs['IWA']

JWSTData assigns a default value for IWA, but run_obs() overwrites this with the user-specified value, defaulting to 1.0 if none is provided.

Currently, this JWSTData configuration is not fully propagated to downstream contrast-analysis routines. For example, calibrate_contrast() and extract_companions() initialized the dataset with only:

pyklip_dataset = JWSTData(filepaths, psflib_filepaths) and
dataset = JWSTData(filepaths, psflib_filepaths, highpass=highpass)

This PR updates these routines so that the downstream JWSTData instances use the same highpass, center_include_offset, center_keywords, and IWA values as the original dataset created in run_obs(). The IWA and HIGHPASS values are also now tracked in the spaceKLIP database and written to the FITS headers.

This PR also adds support for the annuli_spacing parameter in run_obs(). This controls how the KLIP subtraction annuli are distributed radially and can be set to 'constant', 'log', or 'linear'. Previously, annuli_spacing could not be passed through run_obs(), so pyKLIP always used its default 'constant' spacing. The selected value is now stored in the spaceKLIP database as ANNSPACE, and used in downstream routines calibrate_contrast() and extract_companions().

The slides attached to this PR compare the Stage 3 PSF-subtracted products with products generated during contrast analysis after injecting a zero-flux companion. Since a zero-flux injection should leave the data unchanged, we should see no differences when comparing the klipped products. The comparisons include ADI, RDI, and ADI+RDI reductions for both the NIRCam and MIRI demo datasets. The "After" images in the slides show that the same KLIP parameters are now getting passed downstream.
klip_before_after.pptx

When comparing the extract_companions() results from the current develop branch with those produced using the fixes in this PR (using the same set up as in the tutorial notebooks), the best-fit companion parameters show small differences. Across the fitted parameters, the median percent difference is ~ 0.025% for NIRCam and 0.43% for MIRI params. The largest percent differences are generally associated with the uncertainty (*_ERR) values.

This should address the concerns brought up in the following other issues: #148 and #305

This PR also includes a small edit to plotting.py, since get_fits_wcs has been deprecated in stdatamodels. This update is needed for the plotting routines used in the tutorials to run.

[AarynnCarter]

Hey Kayli,

This looks great - thanks for taking a look! My only major comment is that we're sort of jumping over ourselves to assign the IWA value. If you look at the pyKLIP for JWSTData here:

https://github.com/semaphoreP/pyklip/blob/5f6289093d20d41950095eedffd1b3ab2fb9db1a/pyklip/instruments/JWST.py#L290

You'll see that we have an automatic calculation for the IWA based on the wavelength and mode being used, and we don't want to replace it entirely with a separate hardcoded value in spaceKLIP. Instead, I'd suggest we leave the IWA as is unless the user provides a value. Perhaps we could have an 'auto' setting that is the default which runs through pyKLIP, but if the code sees a float or an integer there then it knows it needs to replace the default pyKLIP value?

The other alternative is we edit pyKLIP and pull the IWA out entirely, but I think there's some value to keeping it in there.