Input parameter `dim_input` is not actually used, and always 3000 highly variable genes in function `preprocess`.

[RobinZe]

Expected Behavior:
When initializing the GraphST class with an adata without being preprocessed, the preprocessing step should calculate the top dim_input highly variable genes to match the model's expected input dimensionality.

Actual Behavior:
The preprocess function hardcodes the n_top_genes parameter to 3000, regardless of the dim_input value provided. This results in the model being initialized with a dim_input that doesn't correspond to the number of highly variable genes selected, leading to a mismatch in dimensions.

Relevant Code Snippet:

def preprocess(adata):
    sc.pp.highly_variable_genes(adata, flavor="seurat_v3", n_top_genes=3000)
    sc.pp.normalize_total(adata, target_sum=1e4)
    sc.pp.log1p(adata)
    sc.pp.scale(adata, zero_center=False, max_value=10)

When I tried to assign dim_input = 1500:

model = GraphST.GraphST(adata, datatype='Stereo', device=torch.device('cuda:0'), dim_input=1500)

I also got 3000 HVGs:
'model.dim_input

3000
model.adata.obsm['feat'].shape
(21103, 3000)`

Proposed Solution:
Modify the preprocess function to use the dim_input parameter when selecting the top highly variable genes. Here's the corrected code:

def preprocess(adata, dim_input):
    sc.pp.highly_variable_genes(adata, flavor="seurat_v3", n_top_genes=dim_input)
    sc.pp.normalize_total(adata, target_sum=1e4)
    sc.pp.log1p(adata)
    sc.pp.scale(adata, zero_center=False, max_value=10)