Consider dropping hash-based identification

[selkamand]

Now that we've added #47 and attribute based identification to sig_identify_collection(), should we drop the digest-package dependent hash based identification?

Pros:

• Drops a dependency (digest)
• Reduces the number of tests required (while it is an option we need to test sig_identify_collection() in hash-only mode to make sure the reference db is always up to date).
• Reduces code complexity.

Cons:

• If a signature collection is stored as a csv file and read into R using read.csv. The hash based method is theoretically capable of identifying when it matches a sigverse collection without any special attributes being added. Its unclear at this stage how important this feature will be (importance could be mitigated by adding a parse_signature_collection() function that automatically adds the relevant attributes e.g. collection_name = filename ).
• The digest package depends only on utils - so its a VERY light package and not a major problem for dependency identification
• More robust to identify collections based on content rather than metadata (at least now that locale-dependence of sorting pre-md5 computation has been removed)

[selkamand]

We already have:

sigstash::sig_read_signatures()
sigstash::sig_write_signatures()

We should add attributes addition to sig_read_signatures()