pex3 lock sync does not work with multi-resolve universal lockfile + pytorch

Using latest version of Pex (`2.96.1`), given a simple requirement set with just `pytorch`: ``` torch>=2.9.1,<3.0.0; sys_platform != 'darwin' torch>=2.9.1,<3.0.0; sys_platform == 'darwin' ``` (note the duplicated requirement is to force a separate resolve for MacOS) Using this command to create a universal lockfile with pytorch's CPU wheels: ``` pex3 lock create \ --requirement requirements.txt \ --index "https://pypi.org/simple/" \ --index "pytorch=https://download.pytorch.org/whl/cpu" \ --source "pytorch=torch" \ --indent 2 \ -o default.lock \ --style universal \ --interpreter-constraint "==3.12.*" ``` Resolution completes successfully, and as expected the resulting lockfile has two resolves, one for MacOS with `torch==2.12.1` and one for the rest with `torch==2.12.1+cpu`. However, it is now not possible to update this lockfile with `pex3 lock sync`, as even with _no change at all_ (so it should be a no-op) it fails: ``` pex3 lock sync \ --requirement requirements.txt \ --index "https://pypi.org/simple/" \ --index "pytorch=https://download.pytorch.org/whl/cpu" \ --source "pytorch=torch" \ --indent 2 \ --lock default.lock \ --style universal \ --interpreter-constraint "==3.12.*" ``` ``` ERROR: Given the lock requirements: torch<3.0.0,>=2.9.1; sys_platform != "darwin" torch<3.0.0,>=2.9.1; sys_platform == "darwin" The following lock update constraints could not all be satisfied: cuda-bindings==13.3.1 cuda-pathfinder==1.5.5 cuda-toolkit==13.0.2 filelock==3.29.4 fsspec==2026.6 jinja2==3.1.6 markupsafe==3.0.3 mpmath==1.3 networkx==3.6.1 nvidia-cublas==13.1.1.3 nvidia-cuda-cupti==13.0.85 nvidia-cuda-nvrtc==13.0.88 nvidia-cuda-runtime==13.0.96 nvidia-cudnn-cu13==9.20.0.48 nvidia-cufft==12.0.0.61 nvidia-cufile==1.15.1.6 nvidia-curand==10.4.0.35 nvidia-cusolver==12.0.4.66 nvidia-cusparse==12.6.3.3 nvidia-cusparselt-cu13==0.8.1 nvidia-nccl-cu13==2.29.7 nvidia-nvjitlink==13.0.88 nvidia-nvshmem-cu13==3.4.5 nvidia-nvtx==13.0.85 setuptools==81 sympy==1.14 torch==2.12.1+cpu triton==3.7.1 typing-extensions==4.15 Encountered 1 error updating default.lock: 1.) cp312-cp312-manylinux_2_39_x86_64: pid 3769 -> /home/tmarback/.cache/pex/venvs/3/c6117f5843d015aabb1f1ab35ff7b68a02dda062/0de1795ad4486f45ee94cecd983e9905b6b11dc9/bin/python /home/tmarback/.cache/pex/venvs/3/c6117f5843d015aabb1f1ab35ff7b68a02dda062/0de1795ad4486f45ee94cecd983e9905b6b11dc9/pex --disable-pip-version-check --exists-action a --no-input --no-python-version-warning --isolated --log /tmp/pex-pip-log.ocb2zz2e/pip.log -q --cache-dir /home/tmarback/.cache/pex/pip/3/23.2/pip_cache install --no-clean --dry-run --ignore-installed --report /home/tmarback/.cache/pex/downloads/1/.tmp/resolver_report.i241ac5_/universal-mac-4244e55870f3fe801640657563437be4427eafac/pip-report.json --requirement /home/tmarback/torch-test/pex_lock_split.fqwxxwv9.requirements2.txt --constraint /tmp/lock_update.lkqiyyte.constraints.txt torch<3.0.0,>=2.9.1; sys_platform == "darwin" --index-url https://pypi.org/simple --extra-index-url https://pypi.org/simple/ --retries 5 --timeout 15 exited with 1 and STDERR: pip: ERROR: Cannot install torch<3.0.0 and >=2.9.1 because these package versions have conflicting dependencies. pip: ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts pip: pip: The conflict is caused by: pip: The user requested torch<3.0.0 and >=2.9.1 pip: The user requested (constraint) torch==2.12.1+cpu pip: pip: To fix this you could try to: pip: 1. loosen the range of package versions you've specified pip: 2. remove package versions to allow pip attempt to solve the dependency conflict ``` This does _not_ happen if there is only one entry for `torch` in `requirements.txt` without the environment markers (so only one resolve is generated), but then the lockfile does not work on MacOS. Sidenote, the same happens if `requirements.txt` only has `torch` once but there is a `constraints.txt` with the environment markers instead.