Handle experimental drug names

Drugs are often given company- or target-specific code names, with a prefix followed by an ID number (for example, AstraZeneca uses the prefix AZD for in-development therapeutics -- their COVID vaccine was originally named `AZD1222`). However, curators inconsistently format these codenames, sometimes using a space (`AZD 1222`), sometimes a dash (`AZD-1222`), and sometimes no separator (`AZD1222`), and our sources may not include each possible form as aliases. We should assemble a list of such prefixes and ID structures and enable a check for matches under each possible form if a user's query looks like it conforms to this structure. This behavior could look similar to the current namespace inferrence that we can perform for some kinds of concept IDs.

Start by examining the current namespace inference for concept IDs, since the issue says the new behavior should be similar. Define the drug-code prefixes and ID structures to recognize, then determine how queries should be checked across spaced, dashed, and unseparated forms. Done means matching queries can find aliases under each supported form.

Written by the indexing model from the issue text.