OIIO 3.0/3.1: nmiplevels set to 0

[pkilgo-blz]

Hi, I was noticing in OIIO 3.0 and 3.1, some internal tests caught that nmiplevels was set to 0 after reading an exr file with miplevels in it. I've reworked into a repro script:

import os
import OpenImageIO as oiio
print("OIIO", oiio.__version__)

SIZE = 8
MAX = 10
spec = oiio.ImageSpec(SIZE, SIZE, 3, oiio.FLOAT)
buf = oiio.ImageBuf(spec)

output_path = "repro.exr"

ok = oiio.ImageBufAlgo.make_texture(oiio.MakeTxTexture, buf, output_path)
assert ok

read_buf = oiio.ImageBuf(output_path)
print(f"nmiplevels = {read_buf.nmiplevels}")

miplevel = 0
while True:
    #ok = read_buf.read(0, miplevel=miplevel, force=True)  # OIIO 2.x
    ok = read_buf.read(0, miplevel=miplevel)  # OIIO 3.0+
    if not ok:
        assert read_buf.geterror()
        break
    w, h = read_buf.spec().width, read_buf.spec().height
    print(f"  miplevel {miplevel}: {w}x{h} [nmiplevels={read_buf.nmiplevels}]")
    miplevel += 1
    if miplevel > MAX:
        break

Here is some sample output I got:

$ py311/bin/python repro-nmiplevels.py 
OIIO 3.1.8.0
nmiplevels = 0
  miplevel 0: 8x8 [nmiplevels=0]
  miplevel 1: 4x4 [nmiplevels=0]
  miplevel 2: 2x2 [nmiplevels=0]
  miplevel 3: 1x1 [nmiplevels=0]

Mostly just trying to get a gut check that I haven't bumped into a deprecation. I compared against OIIO 2.1 and 2.4 which seem to set the nmiplevels to 4 for this case.