read_vrt(band=None) only masks band 0's nodata sentinel on multi-band integer VRTs

[brendancol]

Describe the bug

read_vrt(path) on a multi-band integer VRT with per-band <NoDataValue> tags only masks band 0's sentinel. Bands 1 and up keep their integer sentinels as literal finite values in the returned float64 array.

The float-VRT path masks per-band sentinels correctly in _vrt._read_data (lines 296-297 and 347-351). For integer rasters the masking happens later, in __init__.py:read_vrt (lines 2795-2809), and only uses vrt.bands[0].nodata, so bands 1+ are left untouched.

Same class of bug as #1598. PR #1602 fixed band=N selection (single-band reads of band N). The band=None multi-band case still mis-masks every non-first band.

To Reproduce

import tempfile, os, numpy as np
from xrspatial.geotiff import read_vrt
from xrspatial.geotiff._writer import write

with tempfile.TemporaryDirectory() as td:
    b0 = np.array([[1, 2], [3, 65535]], dtype=np.uint16)
    b1 = np.array([[7, 8], [9, 65000]], dtype=np.uint16)
    p0, p1 = os.path.join(td, 'b0.tif'), os.path.join(td, 'b1.tif')
    write(b0, p0, nodata=65535, compression='none', tiled=False)
    write(b1, p1, nodata=65000, compression='none', tiled=False)

    vrt = os.path.join(td, 't.vrt')
    open(vrt, 'w').write(f'''<VRTDataset rasterXSize="2" rasterYSize="2">
  <GeoTransform>0.0, 1.0, 0.0, 0.0, 0.0, -1.0</GeoTransform>
  <VRTRasterBand dataType="UInt16" band="1"><NoDataValue>65535</NoDataValue>
    <SimpleSource><SourceFilename relativeToVRT="0">{p0}</SourceFilename><SourceBand>1</SourceBand><SrcRect xOff="0" yOff="0" xSize="2" ySize="2"/><DstRect xOff="0" yOff="0" xSize="2" ySize="2"/></SimpleSource></VRTRasterBand>
  <VRTRasterBand dataType="UInt16" band="2"><NoDataValue>65000</NoDataValue>
    <SimpleSource><SourceFilename relativeToVRT="0">{p1}</SourceFilename><SourceBand>1</SourceBand><SrcRect xOff="0" yOff="0" xSize="2" ySize="2"/><DstRect xOff="0" yOff="0" xSize="2" ySize="2"/></SimpleSource></VRTRasterBand>
</VRTDataset>''')

    r = read_vrt(vrt)
    print(r.values[1, 1, 0], r.values[1, 1, 1])
    # Got: nan 65000.0
    # Expected: nan nan

Expected behavior

Every band's per-band <NoDataValue> sentinel should be NaN-masked in the returned float64 array, matching the float-VRT path.

Additional context

• HIGH severity: silent wrong-but-finite output for nodata pixels in bands 1+. Downstream code that uses np.isnan for nodata detection will treat 65000, -9999, etc. as real data.
• Related: read_vrt(band=N) reports band 0's nodata in attrs and skips masking for band N #1598 (PR Fix read_vrt(band=N) using band 0's nodata sentinel (#1598) #1602) fixed the band=N single-band selection path.
• The attrs['nodata'] value is intentionally band 0's sentinel for band=None reads (one attr can't encode per-band values); this issue is about the pixel mask, not the attr.
• Fix scope: in __init__.py:read_vrt, when band is None, loop over vrt.bands and mask each band's slice against its own <NoDataValue>, gated by the existing _int_nodata_in_range helper from PR Fix OverflowError on uint TIFF with negative GDAL_NODATA sentinel (#1581) #1583.