Improve detection of boundaries

[Aclrian]

Describe the bug

Right now, not all boundaries are detected. Thus, NamedTypes of parameters include not only "int" or "float", but also the boundary. That makes NamedType hard to use.

To Reproduce

parse scikit-learn to api.json
or
look take a look at sklearn__api.json

Expected behavior

The NamedType should only include the type of parameters and not information about the boundary.

Screenshots (optional)

No response

Additional Context (optional)

Examples of NamedTypes that include a boundary:

int >1
float > 0 and <= 1
float between 0 and 1
float in [0.0, 1.0]
float ([0, 1])
float within (0.0, 1.0)
float in range [0.0, 1.0]

In _migrate_value_annotation.py is "int " and "float " used to include these cases. These conditions can be removed, if this is fixed.