[BUG] Gene equality outside of models

[oxinabox]

Hi,
Several bugs around the handling of genes in reactions when not connected to models
which were found by @jedforrest.
Mostly they boil down to genes with same ID not always being treated as identical.
I think they probably should be, since e.g. gapfilling treats reactions of same ID as identical.
And once they are in a model then the same ID is treated as identical, for perventing duplicates being added when adding a series of reactions.

Declared seperately in GPR from string not equal

import ast
from cobra.core.gene import GPR
from cobra import Model, Reaction

def print_genes(genes):
    print(sorted(genes, key=lambda x: x.id))

# updating GPR by string
rxn1 = Reaction('rxn1')
rxn1.gene_reaction_rule = "(A and B) or C"
genes_abc = rxn1.genes
rxn1.gene_reaction_rule = "A and B"
genes_ab = rxn1.genes

Results in:

In [2]: genes_ab.issubset(genes_abc) # False!
Out[2]: False

In [3]: print_genes(genes_abc)
   ...: print_genes(genes_ab)
[<Gene A at 0x7e3f246f2a80>, <Gene B at 0x7e3f15f2bce0>, <Gene C at 0x7e3f16a8d460>]
[<Gene A at 0x7e3f98e1fe60>, <Gene B at 0x7e3f99cf0fb0>]

Created via GPR manipulation:

This one might arguably need model.repair called to fix it.
But this isa Reaction so there is no such model.

# updating GPR by AST Expression
rxn2 = Reaction('rxn2')
rxn2.gene_reaction_rule = "(A and B) or C"
genes_abc = rxn2.genes
sub_op = rxn2.gpr.body.values[0]  # '(A and B)'
rxn2.gpr = GPR(ast.Expression(sub_op))
genes_ab = rxn2.genes

similarly gives

In [5]: genes_ab.issubset(genes_abc) # False!
Out[5]: False

In [6]: print_genes(genes_abc)
   ...: print_genes(genes_ab)
[<Gene A at 0x7e3f1b5c33e0>, <Gene B at 0x7e3f1682f620>, <Gene C at 0x7e3f1930d310>]
[<Gene A at 0x7e3f16a8d460>, <Gene B at 0x7e3f15f2bce0>]

adding a reaction to a creates new nonequal genes

model = Model("model")
rxn3 = Reaction('rxn3')
rxn3.gene_reaction_rule = "(A and B) or C"
rxn3_genes_orig = list(rxn3.genes)

model.add_reactions([rxn3])

In [8]: 
   ...: rxn3.genes == rxn3_genes_orig  # False!
Out[8]: False

In [9]: print_genes(rxn3_genes_orig)
   ...: print_genes(rxn3.genes)
[<Gene A at 0x7e3f1682e5a0>, <Gene B at 0x7e3f998235c0>, <Gene C at 0x7e3f1a7304a0>]
[<Gene A at 0x7e3f98e485f0>, <Gene B at 0x7e3f98e482f0>, <Gene C at 0x7e3f98e481d0>]

---

I have seen a situation, though I don't have a MWE where I ended up with duplicate genes in the same model.Genes with the same id.
Which breaks the DictList.
But I don't have a small reproducer yet.