[CI] Adjust for decommissioned runners (#24951)

Adjust for decommissioned runners:

 - Disable "Test AMD W7900 / test_w7900"
- Allow "Test ONNX / test_onnx_models :: amdgpu_vulkan_rdna3" on other
runners

.github/workflows/pkgci_test_amd_w7900.yml

diff --git a/.github/workflows/pkgci_test_amd_w7900.yml b/.github/workflows/pkgci_test_amd_w7900.yml
index ceda47c..c5c9762 100644
--- a/.github/workflows/pkgci_test_amd_w7900.yml
+++ b/.github/workflows/pkgci_test_amd_w7900.yml
@@ -19,6 +19,8 @@
 
 jobs:
   test_w7900:
+    # Disabled until an iree-w7900 runner is available.
+    if: false
     runs-on: [Linux, X64, iree-w7900]
     env:
       PACKAGE_DOWNLOAD_DIR: ${{ github.workspace }}/.packages

.github/workflows/pkgci_test_onnx.yml

diff --git a/.github/workflows/pkgci_test_onnx.yml b/.github/workflows/pkgci_test_onnx.yml
index 106890c..3569537 100644
--- a/.github/workflows/pkgci_test_onnx.yml
+++ b/.github/workflows/pkgci_test_onnx.yml
@@ -159,11 +159,7 @@
             runs-on: [Linux, X64, gfx1201, persistent-cache]
           - name: amdgpu_vulkan_rdna3
             config-file: onnx_models_gpu_vulkan.json
-            # TODO(#24464): Remove `shark10-ci` label. There are vulkan driver
-            # issues on shark55. Note that the test is passing on shark01 as
-            # well. shark10-ci is picked because it's what the previous config
-            # was.
-            runs-on: [Linux, X64, rdna3, persistent-cache, shark10-ci]
+            runs-on: [Linux, X64, rdna3, persistent-cache]
 
           # NVIDIA GPU
           # TODO(#18238): migrate to new runner cluster