Fix proof of non-inclusion in SMT Proof Verification algorithm

[parasyte]

This was discussed in #319 (comment)

Adds support for non-inclusion proofs to the algorithm description and pseudocode. Fixes some minor syntactic errors (missing semicolons) and slightly improves readability with whitespace.

The cachedZero pseudocode was also updated to use n as the index variable name to better match the longer pseudocode.

And today I learned that mdBook's smart-punctuation option allows rendering em dashes with triple-hyphens. https://rust-lang.github.io/mdBook/format/markdown.html#smart-punctuation Em dashes looked better than parentheses and commas.

[parasyte]

Another change I made that I forgot to note: The bit-indexing was replaced with a .bitAt() method. I can't explain why I used camelCase, but there it is. I can change it to snake_case if preferred. I think the method's intention is clear without additional commentary, but maybe it needs some extra exposition.

[jintekc]

on a separate and unrelated note. I think nonce should be REQUIRED, not OPTIONAL in data-structures.md. Both leaf-value branches compute hash(hash(proof.nonce)), so every proof needs the nonce. Maybe we could sneak that into this PR as well.

[danpape]

on a separate and unrelated note. I think nonce should be REQUIRED, not OPTIONAL ...

Hmm, yeah, we did write it like

nonce: OPTIONAL 256-bit nonce generated for each update. MUST be encoded as a string ...

I sure don't remember whatever discussion we had last year about why this might be optional or not, but it does seem like we kind of require it now.

[parasyte]

nonce was made optional because it bloats proofs when the additional privacy isn't needed. A proof of non-inclusion is slightly smaller without a nonce, and as the number of aggregate update rounds increases, it can eventually create a huge number of proofs, almost all of them non-inclusions.

@rxgrant had some thoughts on this, specifically.

I don't have a problem with it either way. The nonce is not required for security. It increases privacy by making all proofs (including non-inclusion) indistinguishable from proofs of inclusion. Whether that privacy is useful is an end-user decision.