Fix error when using an external clustering

[JeanMainguy]

This PR addresses a few issues that users encountered when running the cluster command with an external cluster file.

Fixes made:

• Avoid unnecessary gene sequence loading: Previously, gene sequences were being requested and loaded even when using an external cluster file. This was a problem if the pangenome didn’t have sequences.
• Ensure correct data types for gene and family IDs: When gene IDs were purely numeric, pandas was reading them as integers, causing mismatches since gene and family IDs should be treated as strings. We now explicitly ensure these columns are read as strings, avoiding any confusion.
• Clear error for missing genes: If a gene from the cluster file wasn’t found in the pangenome, the code didn’t raise an error immediately. This led to confusing errors later. Now a clear error is raised.
• Clear error for singleton: When a singleton family is created and that a family with the same name is already contained in the pangenome, an unclear error was raised. Now the error states clearly the problem.
• Fix inconsistencies in documentation: The column order for external clustering described in the documentation was inconsistent with the implementation in the code, as highlighted in issue Can't use external clustering: "Exception: Representative gene has not been set" #279. This is now fixed. The correct order is: family_id gene_id representative id. The families_tsv documentation was also out of date.

[axbazin]

In the documentation, we do not discuss the case where we provide 3 columns where it is "family_id gene_id fragmentation" in the doc. Is this voluntary and we have it only to support the old format, or do we want users to use this possibility of giving external clusterings with the info of fragmentation, but without the info of family representative?

[axbazin]

since we're in a "read external cluster' case, shouldn't we just not load gene sequences even if there are some?

[JeanMainguy]

This is used to add protein sequences of representative genes at the end.
It's used when context and fasta commands are apply to the pangenome I believe.
Since pangenomes made with internal clustering already contain these sequences, I guess it is better to also have them in pangenomes generated with external clustering when it is possible to ensures that both types of pangenomes behave consistently...

[JeanMainguy]

I updated the documentation to clarify that we are also taking cluster file made of 3 columns with : famille, gene and fragmentation.