We have only just started moving into super-pangenomes which include species within a genus, typically a cultivated variety with it's wild relatives. See a very important perspectives paper here.

You are suggesting mixing considerably different genomes into a single accession. Just to highlight some of the differences you would need to overcome, the wheat genome has ~100,000 coding genes and is 14.5Gb. Rice has ~36,000 coding genes and is 0.4Gb.

In our experience, it's hard enough to compute whole genome alignments (WGA) among assemblies from the same species to even consider wider taxonomic levels.

If you really want to do something like that you might want to construct a gene-level pangenome. This approach can be done without WGAs and can be computed from nucleotide or protein sequences. Nucleotides give you more resolution (see GET_HOMOLOGUES-EST), proteins allow you to increase evolutionary distances and combined with gene order have been shown to work across species (see for instance GENESPACE).