Speed up `is_dir` on entries returned by `iterdir` and `glob`

The `is_dir` check is fairly expensive, but at least for S3 and Azure when the entries were created as a result of the client's `_list_dir` method, you can tell for each entry whether it is a directory or a file and immediately set the result on the created CloudPath instance.

For example for the `S3Client._list_dir`, you could write something like:

```python
paginator = self.client.get_paginator("list_objects_v2")

for result in paginator.paginate(
Bucket=cloud_path.bucket, Prefix=prefix, Delimiter="/", MaxKeys=1000
):

# sub directory names
for result_prefix in result.get("CommonPrefixes", []):
path = S3Path(f"s3://{cloud_path.bucket}/{result_prefix.get('Prefix')}")
path._is_dir = True
yield path

# files in the directory
for result_key in result.get("Contents", []):
path = S3Path(f"s3://{cloud_path.bucket}/{result_key.get('Key')}")
path._is_dir = False
yield path
```

and modify `S3Path.is_dir`:

```python
def is_dir(self) -> bool:
if self._is_dir is None:
self._is_dir = self.client._is_file_or_dir(self) == "dir"
return self._is_dir
```

This makes a HUGE performance difference if you need to call `is_dir` on the entries returned from `iterdir` or `glob` (in my case, when implementing a file dialog that works for cloud paths).

Not sure if this particular implementation is the best way to do this, but something like this is needed.