GFA import - block group naming

[bobvh]

GFA Import: Sequence Graph Name Not Derived from File (CLI & Python)

Problem

When importing a GFA file (via CLI or Python bindings), the sequence graph (aka block group) is always created with an empty name (""), unlike FASTA and GenBank imports which use the entry name from the file.
Comparison:

Import Type	CLI Block Group Name	Python Block Group Name
GFA	Always ""	Always ""
FASTA	From file (sequence name/ID)	From file
GenBank	From file (locus name)	From file
Library	CLI arg library_name	Arg library_name

Relevant code locations:

• CLI: src/commands/import/gfa.rs:52 - passes "" to BlockGroup::create
• Python: gen-python/src/imports.rs:80 - calls CLI function with empty name

Impact

This affects both CLI and Python users:

• All GFA-imported block groups have the same (empty) name
• Difficult to identify or query specific GFA imports afterward
• Inconsistent behavior compared to other import types
• Importing the same GFA file multiple times keeps adding more nodes and edges to the same object

Short-term Solution

Use the filename without extension as the default block group name, with an optional parameter to override it.
CLI changes (src/commands/import/gfa.rs):

1. Add new CLI argument: --graph-name (optional) to explicitly set the block group name
2. If not provided, derive from filename: Path::new(&path).file_stem().to_string_lossy()
   Python bindings (gen-python/src/imports.rs):
3. Add optional graph_name: Option<String> parameter to import_gfa()
4. Pass through to Rust function; derive from filename if not provided

Example Usage

CLI:

# Uses "my_design" as block group name (derived from filename)
gen import gfa my_design.gfa --sample s1
# Explicitly overrides block group name to "custom_name"
gen import gfa my_design.gfa --sample s1 --graph-name custom_name

Python:

# Uses "my_design" as block group name (derived from filename)
db.import_gfa("my_design.gfa", sample="s1")
# Explicitly overrides block group name to "custom_name"
db.import_gfa("my_design.gfa", sample="s1", graph_name="custom_name")

Related: Larger model renaming effort

The current --name flag for collection is confusing. What I'm hoping we conclude there will have us treating collections like filesystem paths with analogous navigation concepts (nesting, cd, pwd, ls). Reserving --name for the sequence graph name seems sensible, provided we have a good way to deal with multiple contigs per file.