fix: pass links to store when adding data

[rklaehn]

Note that this may lead to decreased performance, since we are doing more work than before. But the store needs this information.

The way links are parsed can be optimized, see n0-computer/beetle#162 . And I am not quite sure if finding links should not be done on the store side. But that is another discussion to be had.

Fixes n0-computer/beetle#167

[rklaehn]

I am now passing through all the links, so at all places where there was a (Cid, Bytes) before there is a (Cid, Bytes, Vec> now. But I don't like it very much. It is a bit verbose and error prone. So will try making a struct and see how that goes...

[rklaehn]

@dignifiedquire had a bit of a fun time merging this with the changes from @ramfox in 6ba6b57 , but I think it turned out OK.

Regarding the tests: I have added validate to block, but I am not quite sure where to put the tests. I can easily validate that the blocks that are produced by file.encode or dir.encode are correct by adding a call to validate in stream_to_resolver, which is used by the roundtrip tests.

But I think what you might want is to actually test that links are properly passed to the store in add_file or add_dir, which would be big enough to do a separate issue. I think there is currently no test at all for unixfs_builder::add_dir or unixfs_builder::add_file. And this would have to be more like an integration test that creates a temp dir etc.

If you want I can add it to this PR, but it will take some time.

Edit: added the former because why not.

[dignifiedquire]

Lets get this in and make a new issue for adding adding tests

[rklaehn]

Made a followup issue for the proposed tests: n0-computer/beetle#152

[dignifiedquire]

nice catch