chore: auto-open resources

[gmaclennan]

This removes some of the changes in #1232 around resource opening. Putting initialization in _open() creates an additional state to manage for classes, "initialized but not open", and the need to wrap some functions and methods in a this.ready() guard, and a condition for a future dev "you must sometimes wait this.ready() before you access some properties". For example, #1232 was missing guards for BlobStore.get(), createReadStream(), createEntriesReadStream(). This isn't currently well covered in the tests because most/all reads happen after opening and writing resources - we're lacking tests that initialize a new instance with existing data and immediately try to read it.

There's no need for the async open setup anyway. All of the upstream hypercore modules wrap their ready state, so that you can call any method before they are ready. The CoMapeo Core code has an intentional architecture that removes the need to access hypercore properties that are not populated until after ready() e.g. core.key.

This adds a — currently unnecessary — call to this.ready().noop() in the constructor of all classes that use ReadyResource. I've done this because not all of our methods call this.ready() first so we don't guarantee a lazy init, but it works, for now, because the upstream holepunch modules either lazy-init or auto-open. I think it's safer to have a "init in constructor" pattern rather than try to implement lazy init, because lazy init opens too many opportunities for bugs down the line with little gain.

[gmaclennan]

Now that tests are using random-access-file, they are running much slower than with random-access-memory, and occasionally run more than 10 mins, which was our previous CI timeout value. This PR also increases the CI timeout to 15 minutes.