ChEBI-based curation of metabolite names

[edkerk]

Split from #491, where the specific typo that prompted it (MAM02053) was fixed in #518. This issue tracks the broader idea from that thread: curating metabolite names against ChEBI.

Goal

Compare Human-GEM metabolite names with the names of their assigned ChEBI identifiers, and correct clear discrepancies.

Notes from the earlier discussion

Not every mismatch should be changed:

• Many differences are stylistic and the model name is fine to keep (for example "O2" vs "dioxygen", or stereo-notation differences like "18-(R)-" vs "(18R)-").
• The ChEBI preferred name is not always the best choice, and its synonyms are often equally valid. For example CHEBI:39248 is "henicosanoic acid" in ChEBI, but the model consistently uses the synonym "heneicosanoic acid"; internal consistency across related metabolites matters more than matching the ChEBI preferred string.
• Some "pool" metabolites (modeling abstractions) carry ChEBI IDs and should be reviewed separately, since they are not single chemical entities.
• Some ChEBI IDs themselves may need correcting, not only the names.

Suggested approach

1. Regenerate the model-name vs ChEBI-name diff against the current model/metabolites.tsv (the 2023 analysis attached in Typo in the name of MAM02053 #491 predates v2.0.0 and is stale). Add it as a TSV under data/modelCuration/.
2. Apply the clear, unambiguous corrections first (genuine typos and errors), leaving stylistic differences alone.
3. Handle pool metabolites and any ChEBI ID corrections as separate follow-ups.

[edkerk]

One cross-dependency worth noting here: since #1052, most reaction names are generated from metabolite names. Transport names embed the transported metabolite, the EC- and gene-based names carry the primary substrate, and the chemistry-derived names embed the substrate and product (for example "fructose phosphorylation" or "ascorbate to monodehydroascorbate conversion").

So correcting a metabolite name in this issue will leave any reaction name that embeds the old string inconsistent. The reaction-name generation is scripted, so the cleanest order is to apply the metabolite-name corrections first and then re-run it (or update just the affected reactions), keeping the two in sync.