speed-up by early pruning based on markers

• I have searched the issues of this repo and believe that this is not a duplicate.
• I have searched the documentation and believe that my question is not covered.

Feature Request

As of now poetry spends a lot of time on overrides which doesn't match the environmental version or platform requirements.
They are filtered in the very end of the process, after writing possible scenarios to the lock file. It would be great to have early pruning of such overrides, particularly as more and more packages go for conditional logic in their requriements.
The issue fits the broader call for better support of markers, such as #4670

Consider this example run within a virtual environment with Python 3.8

[tool.poetry.dependencies]
python = "~3.8"
pandas = [
    {version="1.3.*",markers="python_version=='3.9.*'"},
    {version="1.2.*",markers="python_version=='3.7.*'"},
    {version="1.1.*",markers="python_version=='3.10.*'"},
]
....

Petry correctly concludes the empty set of matching packages, however it splits the graph in 6 branches (due to explicit version split and one more implicit in pandas numpy dependency) - and this escalates the more markers we have. It would help to compare the markers early and ignore some of the branches.

NOTE: this is a broader call for being able to support restricted target environments and narrow the lock to prescribed architectures, python versions and so on.

Reproduce the example with Python 3.8 and trace how marker-incompatible overrides create resolver branches and lock-file scenarios. The work is done when incompatible branches are pruned before scenario generation while the empty matching result remains correct; the issue does not name specific files or tests.