[Fix] Wallet import fails or decodes the wrong key when mnemonic words are capitalized

[roomote-conceal-network]

​Created by Roomote. Follow up by mentioning @roomote, in the web UI, or in Discord.

What changed

mnemonicToPrivateKey now lowercases the entire phrase once at the decode entry point, before any validation or decoding happens. Previously, word validation lowercased a throwaway copy while the checksum derivation and word-list index lookup used the original casing, so a capitalized word (e.g. autocapitalized "Iodine" from a mobile keyboard) made the phrase either fail to decode or decode through a missed word-list lookup that produced wordlist-size indexes and garbage key bytes.

Why this change was made

A user retyping their 25-word phrase on a mobile keyboard with autocapitalize could end up importing nothing or — worse — a valid-looking but wrong private key, with no clear error pointing at the casing. Normalizing case once up front makes word validation, checksum verification, and index lookup all operate on identical input, so any casing variation of a valid mnemonic decodes to the same key it did before.

Impact

Mixed-case mnemonic phrases now decode to exactly the same private key as their lowercase originals, instead of failing or silently decoding a different key. Already-lowercase phrases are unaffected (tolower is a no-op), and rejection behavior is preserved: unknown words and tampered checksums are still rejected exactly as before.

Verified with a standalone g++ harness compiled against both the pre-fix and fixed sources: pre-fix, every mixed-case variant (all-capitalized and single-capitalized word) failed to decode to the correct key; post-fix all variants decode correctly and invalid phrases are still rejected. npm run lint, npm run check, and npm run cpp pass. This repo has no JS test suite, and a full native build requires a host RN app, so those were not run.

[roomote-conceal-network]

No new code issues found. See task

_{Reviewed cd67814}