Excessive memory usage with attachments

Hi all

When using the library to parse a existing message with `messsage = Mail.read("test.eml")` and then accessing the parts or attachments objects like `mesage.parts.size` or `message.attachments.size` we see excessive memory usage.

It seems like the attachments are being held fully in memory - the memory usage is very high.
For example with a message having a single attachment, originally 110 MB in size, base64-encoded in the message and thus the total size of the message coming up to 150 MB we see a overhead of memory usage of about factor two - the simple example above uses about 380 MB of memory.

Is there any option to parse a message without loading everything into memory upfront? Like in a streaming fashion?

Start at Mail.read("test.eml") and trace what happens when parts or attachments are accessed. Compare the message size with the roughly 380 MB memory usage described, then determine whether a streaming parse can avoid loading attachment contents upfront; done means the reported excessive memory usage is addressed without breaking parts or attachments access.

Written by the indexing model from the issue text.