bot reintroduces unnecessary cross-compilation for linux-aarch64

[h-vetinari]

I've seen this on a couple of different feedstocks (which had already been shifted to the new native GHA runners) that bot PRs will reintroduce cross-compilation, e.g.

--- a/conda-forge.yml
+++ b/conda-forge.yml
@@ -1,10 +1,8 @@
 bot:
   automerge: true
 build_platform:
+  linux_aarch64: linux_64
   linux_ppc64le: linux_64
-provider:
-  linux_aarch64: default
-  osx_arm64: default
 conda_build:
   error_overlinking: true
   pkg_format: '2'
@@ -12,17 +10,20 @@ conda_forge_output_validation: true
 github:
   branch_name: main
   tooling_branch_name: main
+provider:
+  linux_aarch64: default
+  osx_arm64: default
 test: native_and_emulated

in conda-forge/google-cloud-cpp-feedstock@78b17e4.

We should not be adding cross-compilation for aarch anymore.

CC @jaimergp @mgorny

[mgorny]

Working on it.

[mgorny]

Okay, I suppose this is the CrossCompilationForARMAndPower minimigrator's doing. I suppose we want to disable the ArchRebuild migrator as well, though, and ultimately replace it with one that enables native builds.

#6251 is still pending, though.