MAINT: Guard type checking imports in TYPE_CHECKING block

[slangeveld]

Resolves #725

Imports that are only used to annotate types in static type-checking, are moved into a type-checking block:

if TYPE_CHECKING:
    ...

The block ensures that these imports will only be imported when we are doing static type checking with mypy.

NB! As stated in #725 imports that are used by Pydantic at runtime can not be moved to the TYPE_CHECKING block

[mferrera]

LGTM 👍 remember to squash or rebase the fix-up commit before merging

[mferrera]

This is a nice save, in theory. Both pandas and xtgeo can be slow to import (xtgeo also imports pandas), so whenever we can constrain them to being imported under a TYPE_CHECKING guard it's a good thing. You see this mostly during "start-up", just before execution is beginning, but with Python this can be quite noticeable (10s of seconds).

A big speed impediment to import time is Python compiling the interpreted code into byte code. Usually this means that the first import is slow, and the rest of the imports are "cached" in the sense that things have already been located and compiled, but whenever we can reduce the dependency graph we can make that compilation faster.

[mferrera]

I think you should also be able to guard from pydantic_core import CoreSchema.

fmu-dataio/src/fmu/dataio/datastructure/meta/meta.py

Line 519 in 113fcaf

core_schema: CoreSchema,

fmu-dataio/src/fmu/dataio/datastructure/meta/content.py

Line 673 in 113fcaf

core_schema: CoreSchema,

In these cases it is used only as an annotation to a method parameter, which Pydantic does not inspect. This will be a good dependency to hide for type checking only because pydantic-core is generally not meant to be used directly (it is the Rust-with-Python-bindings backend to Pydantic).