Add node bindings

[turt2live]

My bot-sdk is trying to add crypto for encrypted appservices and it's been made clear that trying to use Olm directly is a lot more pain than it should be, particularly for proliferation of encryption implementations.

The major pieces needed would be:

• Basic encryption/decryption
• Fallback keys (and regular OTKs which I assume is part of the basic encryption/decryption)
• Key sharing (with a way to bypass the unverified device restrictions given the usecase)
• Key backup
• Capability to run multiple concurrent Olm accounts within the same process efficiently. This includes de-duping information such as the rooms and device lists across potentially thousands or millions of in-process users.
• Other crypto bits which I forgot and will be needed to ensure reliability

Cross signing and verification are not required at this stage for the bot-sdk as they don't serve a major purpose when the bot is not a human.

[poljar]

Fallback keys (and regular OTKs which I assume is part of the basic encryption/decryption)

Fallback keys are missing, this is #400.

Key sharing (with a way to bypass the unverified device restrictions given the usecase)

Key forwarding is supported though no bypass is supported. You'll need to define exactly who should receive forwarded keys, but this shouldn't be hard to add.

Key backup

This is supported but the API requires a bit of logic on the caller side.

Capability to run multiple concurrent Olm accounts within the same process efficiently. This includes de-duping information such as the rooms and device lists across potentially thousands or millions of in-process users.

This requires a custom store, possibly a PostgreSQL based one.

All of the above is already implemented in the bot-sdk, I was under the impression that it does only the basics for now?

[turt2live]

The bot-sdk does the basics and fallback keys, but in the next month or two it'll be getting support for the other listed things.

[ShadowJonathan]

Shouldn't we put FFI bindings in a sub-crate? I think they're useful, but they could clutter compile times and build files for the main crate.

[turt2live]

probably. I think node bindings also have implications on the actual Rust code, so a new crate is almost certainly desirable.

I'll be taking a closer look at this in about a week.

[poljar]

I forgot that this one already exists, #699 has a list of task now and a better issue number. Closing as a duplicate.