bump vigra to 1.11.1 build 1038 (ilastik#2941)

* bump vigra to 1.11.1 build 1038
   this is the latest build with boost 1.78.0
* Add patched ilastik-forge channel to dev environment
   Problems on windows solving the environment related to the
   spherical_texture plugin. In particular pyshtools comes with a lot of
   baggage, dependency-wise (that we don't need). We have patched version
   of pyshtools on ilastik-forge/label/patched that do not include those.
* pin numpy to <1.23
   deal with deprecations after

.github/workflows/core-conda-bld.yml

@@ -134,7 +134,7 @@ jobs:
         run: |
           set VOLUMINA_SHOW_3D_WIDGET=0
           conda build --test --override-channels ^
-            -c ./pkgs -c pytorch -c nvidia -c ilastik-forge -c conda-forge ^
+            -c ./pkgs -c pytorch -c nvidia -c ilastik-forge/label/patched -c ilastik-forge -c conda-forge ^
             ./pkgs/noarch/%ILASTIK_PACKAGE_NAME%
         # HACK: due to a bug in conda-build need to point to
         # libarchive explicitly.

conda-recipe/meta.yaml

@@ -26,7 +26,7 @@ outputs:
         - setuptools_scm
       run:
         - python >=3.7
-        - numpy >1.12
+        - numpy >1.12,<1.23
         # aiohttp to enable zarr.storage.FSStore
         - aiohttp
         - appdirs
@@ -61,7 +61,7 @@ outputs:
         - tifffile >=2022
         # build 1.11.1=*_1028 on cf is the first to be compatible with numpy>1.19
         # need to bump this manually until there is a true version bump in vigra
-        - vigra
+        - vigra 1.11.1=*_1038
         - xarray !=2023.8.0,!=2023.9.0,!=2023.10.0
         - z5py
         - zarr 2.*

dev/environment-dev.yml

@@ -1,6 +1,10 @@
 channels:
   - pytorch
 #  - nvidia  # uncomment for pytorch with cuda
+  # ilastik-forge/label/patched contains patched versions of pyshtools
+  # with certain dependencies removed that made solving the ilastik
+  # environment impossible / extremely inflexible
+  - ilastik-forge/label/patched
   - ilastik-forge
   - conda-forge
   - nodefaults
@@ -40,7 +44,7 @@ dependencies:
   - tifffile >=2022
   # build 1.11.1=*_1028 on cf is the first to be compatible with numpy>1.19
   # need to bump this manually until there is a true version bump in vigra
-  - vigra 1.11.1=*_1033
+  - vigra 1.11.1=*_1038
   # xarray versions not compatible with numpy 1.21, 2023.08.0 might be, but lost trust
   # 2023.10.1 correctly pins numpy to 1.22 and up
   - xarray !=2023.8.0,!=2023.9.0,!=2023.10.0