If caching fails (e.g. if type doesn't match), still run the computation and allow caching separately.

[orgadish]

When running something that takes a long time, it's very disappointing if there's an error only in the caching at the end. It would be better if:

1. The computation still returned into the object.
2. There was a way to cache an object directly to bypass this issue when it occurs.

Example error

oep_br_umap <- oep_br_basic_log_data |> 
    umap::umap() |> 
    filecacher::with_cache("basic_log_umap", CACHE, type="rds")

Error in value[[3L]](cond) : 
  Error in .cache$set(label, out): Error setting value for key "basic_log_umap".
 Check if the cache type ('rds') is compatible with the data being stored.
In addition: Warning message:
failed creating initial embedding; using random embedding instead