[Feature] Auto Download

[GaoxiangLuo]

Is your feature request related to a problem? Please describe.
Some existing examples' datasets come from the installed python packages, but when it comes to users' own datasets, downloading the data is a necessary step. It's good to have the data download method integrated into the mode folder as part of the flame python library. Meanwhile, the design of flame system recognizes which dataset can be accessed by which cluster node within the firewall. It maybe good to take this aspect into consideration.

Describe the solution you'd like
As users' data can be stored at various places (e.g., AWS S3, GitHub, zenodo, Google Drive, personal host), some may require MD5, APIs in addition to URL. Also, one may take the maximum number of redirection as a parameter to avoid endless redirections.

Describe alternatives you've considered
If the node is spawn at where the data is located, such as deploying flame system on a phone while the training data is also on the same phone, then downloading may be unnecessary.