Misleading error message when (extract-variants :variants <n> <expr>) fails because of a cycle in the egraph

[gussmith23]

Current egglog extraction fails when there are cycles; see #315.

I just hit an issue where extract-variants fails because of #315, but the error message is misleading. See this example:
https://egraphs-good.github.io/egglog/?program=XQAAgADEAAAAAAAAAAAUHMnnVi1HmurH0_nckuChY4vMVTFLL6zeL8XZl8K_7jEMYHKxoZcOqRi-rAaPNzwGu5xY09mjyi9zsJSE_gyvpIPvMVxi1HwGaAJEGRTIUcM8sV5W10gDMJm7bxCZNbR7oN8Bfg-QmLQbeaAo2guZ33M1Sa0t6tn_2y-QAA%253D%253D

In the console, you'll see the error message

panicked at src/extract.rs:97:60:
extract_variants should be called after extractor initialization

However, the problem isn't due to an uninitialized extractor.

When :variants isn't set, then under the hood extract is called instead of extract_variants. In that case, we get a somewhat clearer error message:

panicked at src/extract.rs:67:17:
No cost for Value { bits: 0 }