Narrow >V post-release exclusion to match spec

[notatallshaw]

Similar to #1140 (the < counterpart), found while looking at edge cases for #1119 and #1120.

The spec says:

The exclusive ordered comparison >V MUST NOT allow a post-release of the given version unless V itself is a post release.

The current code uses _base_version to check this, which strips pre, post, dev, and local from both sides. This is too broad when V is a pre-release or dev release, because it treats post-releases of the base release as post-releases of V.

For example, >1.0.dev1 accepts 1.0 but rejects 1.0.post0:

>>> Specifier(">1.0.dev1").contains("1.0")
True
>>> Specifier(">1.0.dev1").contains("1.0.post0")
False

1.0.post0 is a post-release of 1.0, not of 1.0.dev1. Since 1.0 itself is accepted, excluding 1.0.post0 is incoherent.

The fix replaces _base_version(prospective) == _base_version(spec) with _post_base(prospective) == spec, which checks whether the prospective version is actually a post-release of the spec version rather than just sharing its release segment.

[notatallshaw]

This one feels more intuitive to me than #1140, I am much less concerned about the behavior change it introduces, the fact that currently >1.0.dev matches 1.0 but not 1.0.post0 is surprising and not spec compliant.

[notatallshaw]

The tests I'm proposing in #1144 show that this fixes an edge case where the current behavior breaks the monotonic invariant, that given versions v1 and v2 and v2 > v1 then if the specifier >V matches v1 it should also match v2.