[BUG] [CuTeDSL] cute.coalesce doesn't do anything on dynamic shapes

Which component has the problem?

CuTe DSL

Bug Report

Describe the bug
A mode with a dynamic shape doesn't get coalesced into previous modes, even though the stride is static and the previous modes are fully static too (and that's all that's needed to check whether the modes can be coalesced).

Steps/Code to reproduce bug

import torch
import cutlass.cute as cute
from cutlass.cute.runtime import from_dlpack


@cute.jit
def repro(tensor: cute.Tensor):
    layout = cute.make_layout(
        (16, tensor.shape[1] // 128),
        stride=(8, 128),
    )
    expected = cute.make_layout(
        (16 * (tensor.shape[1] // 128),),
        stride=(8,),
    )

    print("input:    ", layout)
    print("coalesced:", cute.coalesce(layout))
    print("expected: ", expected)


tensor = from_dlpack(torch.empty((4096, 8192), device="cuda"))
tensor = tensor.mark_compact_shape_dynamic(mode=1)
repro(tensor)

Output:

input:     (16,?):(8,128)
coalesced: (16,?):(8,128)
expected:  (?{div=16}):(8)

Environment details (please complete the following information):
Version 4.7.0 of CuTeDSL

Start with cute.coalesce and reproduce the issue using the repro function and dynamic tensor shape shown in the report. Compare the current coalesced layout with the expected layout, then verify that the dynamic mode is coalesced when its stride and preceding modes allow it.