gh-103193: Speedup and inline inspect._is_type

[AlexWaygood]

(Using @sobolevn's benchmark script from #103193 (comment).)

Benchmarks on main:

type[Foo] : 74 ± 0 ns
Foo : 177 ± 1 ns
type[Bar] : 74 ± 0 ns
Bar : 177 ± 0 ns
WithParentClassX : 229 ± 0 ns
Baz : 216 ± 0 ns
WithParentX : 266 ± 1 ns
type[Missing] : 210 ± 0 ns
Missing : 219 ± 1 ns
Slotted : 214 ± 0 ns
Method : 178 ± 0 ns
StMethod : 178 ± 0 ns
ClsMethod : 178 ± 1 ns

Benchmarks with this PR:

type[Foo] : 74 ± 0 ns
Foo : 134 ± 0 ns
type[Bar] : 74 ± 0 ns
Bar : 133 ± 0 ns
WithParentClassX : 187 ± 0 ns
Baz : 171 ± 0 ns
WithParentX : 220 ± 0 ns
type[Missing] : 210 ± 1 ns
Missing : 174 ± 0 ns
Slotted : 169 ± 0 ns
Method : 133 ± 0 ns
StMethod : 133 ± 0 ns
ClsMethod : 133 ± 0 ns

This speeds up this isinstance() call 1.25x relative to main:

fromtypingimport*@runtime_checkableclassFoo(Protocol):
a: intb: intc: intd: inte: intf: intg: inth: inti: intj: intclassBar:
def__init__(self):
forattrnamein'abcdefghij':
setattr(self, attrname, 42)
isinstance(Bar(), Foo)

• Issue: Optimise inspect.getattr_static #103193

[carljm]

The improvements keep on coming!

[carljm]

This has equivalent semantics with less code and fewer assignments, but perhaps it muddies the semantics of the klass var a bit (since it will temporarily be "wrong" in the case that obj is already a type.) No strong feelings, whatever you prefer.

Suggested change

	objtype=type(obj)
	iftypenotin_static_getmro(objtype):
	klass=objtype
	klass=type(obj)
	iftypenotin_static_getmro(klass):

[AlexWaygood]

Hmm. Yeah, I see what you mean, but I think I'd prefer to keep the semantics of klass clear here :)

[AlexWaygood]

The improvements keep on coming!

Optimisations will continue until morale improves!