bump NCCL floor to 2.18.1.1, relax PyTorch pin (#218)

Contributes to rapidsai/build-planning#102

Fixes #217

## Notes for Reviewers

### How I tested this

Temporarily added a CUDA 11.4.3 test job to CI here (the same specs as the failing nightly), by pointing at the branch from rapidsai/shared-workflows#246.

Observed the exact same failures with CUDA 11.4 reported in rapidsai/build-planning#102.

```text
...
  + nccl                     2.10.3.1  hcad2f07_0                  rapidsai-nightly     125MB
...
./WHOLEGRAPH_CSR_WEIGHTED_SAMPLE_WITHOUT_REPLACEMENT_TEST: symbol lookup error: /opt/conda/envs/test/bin/gtests/libwholegraph/../../../lib/libwholegraph.so: undefined symbol: ncclCommSplit
sh -c exec "$0" ./WHOLEMEMORY_HANDLE_TEST 
./WHOLEMEMORY_HANDLE_TEST: symbol lookup error: /opt/conda/envs/test/bin/gtests/libwholegraph/../../../lib/libwholegraph.so: undefined symbol: ncclCommSplit
sh -c exec "$0" ./GRAPH_APPEND_UNIQUE_TEST 
```

([build link](https://github.com/rapidsai/wholegraph/actions/runs/10966022370/job/30453393224?pr=218))

Pushed a commit adding a floor of `nccl>=2.18.1.1`. Saw all tests pass with CUDA 11.4 😁 

```text
...
  + nccl                     2.22.3.1  hee583db_1                  conda-forge          131MB
...
(various log messages showing all tests passed)
```

([build link](https://github.com/rapidsai/wholegraph/actions/runs/10966210441/job/30454147250?pr=218))

Authors:
  - James Lamb (https://github.com/jameslamb)

Approvers:
  - Vyas Ramasubramani (https://github.com/vyasr)
  - https://github.com/linhu-nv
  - https://github.com/jakirkham

URL: #218

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -28,7 +28,7 @@ dependencies:
 - librmm==24.10.*,>=0.0.0a0
 - nanobind>=0.2.0
 - nbsphinx
-- nccl
+- nccl>=2.18.1.1
 - ninja
 - numpy>=1.23,<3.0a0
 - numpydoc
@@ -40,7 +40,7 @@ dependencies:
 - pytest-xdist
 - python>=3.10,<3.13
 - pytorch-cuda=11.8
-- pytorch=2.0.0
+- pytorch>=2.0,<2.4.0a0
 - rapids-build-backend>=0.3.0,<0.4.0.dev0
 - recommonmark
 - scikit-build-core>=0.10.0

conda/environments/all_cuda-125_arch-x86_64.yaml

@@ -30,7 +30,7 @@ dependencies:
 - librmm==24.10.*,>=0.0.0a0
 - nanobind>=0.2.0
 - nbsphinx
-- nccl
+- nccl>=2.18.1.1
 - ninja
 - numpy>=1.23,<3.0a0
 - numpydoc

conda/recipes/libwholegraph/conda_build_config.yaml

@@ -17,7 +17,7 @@ doxygen_version:
   - ">=1.8.11"
 
 nccl_version:
-  - ">=2.9.9"
+  - ">=2.18.1.1"
 
 c_stdlib:
   - sysroot

dependencies.yaml

@@ -87,7 +87,7 @@ dependencies:
           - libraft-headers==24.10.*,>=0.0.0a0
           - librmm==24.10.*,>=0.0.0a0
           - nanobind>=0.2.0
-          - nccl
+          - &nccl nccl>=2.18.1.1
     specific:
       - output_types: conda
         matrices:
@@ -216,14 +216,14 @@ dependencies:
     common:
       - output_types: [conda]
         packages:
-          - nccl
+          - *nccl
   test_python:
     common:
       - output_types: [conda]
         packages:
           - c-compiler
           - cxx-compiler
-          - nccl
+          - *nccl
       - output_types: [conda, requirements]
         packages:
           - ninja
@@ -285,13 +285,13 @@ dependencies:
               # If conda-forge supports the new cuda-* packages for CUDA 11.8
               # at some point, then we can fully support/properly specify
               # this environment.
-              - pytorch=2.0.0
+              - &pytorch pytorch>=2.0,<2.4.0a0
               - pytorch-cuda=11.8
           - matrix:
               arch: aarch64
               cuda: "11.8"
             packages:
-              - pytorch=2.0.0
+              - *pytorch
               - pytorch-cuda=11.8
           - matrix:
             packages:
@@ -318,7 +318,7 @@ dependencies:
     common:
       - output_types: [conda]
         packages:
-          - pytorch=2.0.0
+          - *pytorch
           - cpuonly
   clang_tools:
     common: