Replace custom name generator with petname

[tupe12334]

Opened automatically by the moadim routine "Replace custom code with existing lib, or extract to new lib → draft PR / issue (one repo/run) → Slack".

The custom code today

src/name_gen.rs:1-43 (called from src/multi_workspace.rs:37 inside create_multi_workspace()):

const ADJECTIVES: &[&str] = &["bold", "bright", "brave", /* ...20 total */];
const NOUNS: &[&str] = &["turing", "neumann", "lovelace", /* ...20 total */];

pub fn generate_name() -> String {
    let id = Uuid::new_v4();
    let b = id.as_bytes();
    let adj = ADJECTIVES[b[0] as usize % ADJECTIVES.len()];
    let noun = NOUNS[b[1] as usize % NOUNS.len()];
    format!("{adj}_{noun}")
}

It hand-rolls a Docker/Heroku-style adjective_noun random name for each new multi-repo workspace directory, burning a full UUID v4 just to get two random bytes as array indices. Only 20×20 = 400 possible names, and the distribution is uneven since 256 isn't evenly divisible by 20.

Proposed library: petname (docs.rs)

• Purpose-built for exactly this use case — it's the standard crate people reach for to generate Docker/Heroku-style random resource names.
• Bigger, curated word lists and a proper RNG-backed Namer/petname() API instead of UUID-byte modulo.
• Popularity/maintenance: ~1.7M total downloads, ~146K/month, Rename daemon subcommand to scheme, remove install action #3-ranked CLI-utility crate on lib.rs, 54 dependent crates. Latest release 3.1.0, actively maintained.
• License: Apache-2.0 — compatible with this repo's MIT license (the standard MIT/Apache-2.0 pairing used throughout the Rust ecosystem).

What this PR does

• Adds petname = "3" to Cargo.toml.
• Reimplements name_gen::generate_name() on top of petname::petname(2, "_"), keeping the same function signature so the call site in multi_workspace.rs needs no change.
• Updates the two unit tests to check the new word source's shape (two non-empty parts joined by _, and non-degenerate distinctness across draws) instead of asserting against the old hardcoded word lists.
• uuid stays as a dependency — it's used elsewhere in the crate (hooks.rs, opener/*, issue/*) and isn't being removed.

Before/after call site (multi_workspace.rs:37, unchanged by this PR):

let root = workspaces_root.join(name_gen::generate_name());

Verified locally: cargo build succeeds and cargo test --lib name_gen passes (2/2).

Trade-offs

• New direct dependency (petname); dependency weight is modest since uuid's v4 feature already pulls in RNG machinery.
• Output vocabulary changes — bigger/different word lists than the old 20/20 arrays. Purely cosmetic; nothing in the crate parses or asserts on specific adjective/noun values outside the now-updated name_gen tests.
• Cargo.toml denies clippy::pedantic/nursery/cargo and missing_docs — the new code keeps doc comments and avoids .expect() (uses unwrap_or_else instead, since expect_used is denied here) to satisfy those gates.
• Could not run the full cargo clippy --all-targets gate in this environment — it fails on a pre-existing, unrelated allow_attributes_without_reason violation in src/issue/paths.rs that predates this change. Worth a clean clippy run before merge.

Checklist for whoever picks this up

• Run cargo clippy --all-targets --all-features on a clean checkout and confirm no new lint violations from this diff
• Decide on petname version pin / feature flags (default default-rng + default-words used here)
• Skim the new word list output for anything undesirable as a directory name (petname's default lists are curated, but worth a glance)
• Merge Cargo.lock update cleanly with any concurrent dependency changes