This fixes some problems with identifiers due to the changes to SourmashSignature.name in #1179.

This should probably go into 4.0.0...

Brief writeup of my understanding of the issues so far:

• sig.name can be empty if no name is given on the signature, but this then causes problems in both sourmash lca index and LCA_Database.insert(...)
• this also revealed a problem in the reporting on identifiers in sourmash lca index - if the identifiers are identical (in this case '', but in whatever case) then the number of leftover signatures gets mis-reported.

A particularly simple situation where this arises is in the test test_index_empty_sketch_name, where we:

1. sketch some DNA sequences using sourmash sketch dna, without providing a name;
2. try to index them with lca index;
3. sig.name is now empty, which triggers a problem in LCA_Database.insert(...) with duplicate identifiers;
4. the number of unused identifiers tracked in record_no_lineage is now miscounted as 1 instead of 2, because we used a set.

This PR fixes the following issues:

• sourmash sketch and sourmash compute no longer create empty signatures from empty files and stdin;
• sourmash sketch and sourmash compute set sig.filename to empty string when filename is -;
• in sourmash lca index, record_no_lineage is now a list instead of set;
• in sourmash lca index, ident is taken from sig.filename if sig.name is empty.
• in LCA_Database, ident defaults to str(sig)

TODO:

• XX add tests for new sketch filename behavior on stdin (punted to adjust screed to allow replacement of stdin input #1348)
• add direct lca_db.insert(...) tests
• add test for new sourmash sig describe behavior

Checklist

• Is it mergeable?
• make test Did it pass the tests?
• make coverage Is the new code covered?
• Did it change the command-line interface? Only additions are allowed
  without a major version increment. Changing file formats also requires a
  major version number increment.
• Was a spellchecker run on the source code and documentation after
  changes were made?