Lint the test targets, which clippy currently skips

[kridaydave]

Problem

.github/workflows/ci.yml:29 ran cargo clippy -- -D warnings. That is the same target selection as cargo check, which covers the lib and bin targets only. It never compiled tests/cli.rs (843 lines) or any #[cfg(test)] mod tests block in src/, so all of the test code was unlinted. Cargo.toml has no [lints] table and there is no clippy.toml, so nothing else covered it. cargo fmt --check on the line above does walk test code, so formatting was enforced there and lints were not.

Change

Two flags. --all-targets covers the test targets. --locked matches the flag cargo test on line 32 already uses, so clippy lints the graph CI actually builds rather than a re-resolved one.

The single warning that --all-targets surfaced is fixed in the same change, because -D warnings would otherwise fail the build:

src/fabric.rs:234  assert_eq!(loaded.require_check_strict, false)  ->  assert!(!loaded.require_check_strict)

Verification

I ran the new command on this tree, unmodified otherwise:

$ cargo clippy --all-targets --locked -- -D warnings
    Checking taproot v0.0.1 (/root/code/_watch-worktrees/taproot)
    Finished \`dev\` profile [unoptimized + debuginfo] target(s) in 10.89s

Exit 0, zero warnings. Before the assert! fix, the same command reported exactly one warning, bool_assert_comparison at src/fabric.rs:234, confirming the coverage gap was real rather than theoretical. cargo fmt --all -- --check is clean.

Not shipped

• src/mount.rs:437 casts a FUSE setattr size (u64) to usize and resizes a buffer with it. A truncate through the mount with a size near u64::MAX reaches a multi-exabyte allocation. I did not reproduce it, so it is a write-up rather than a fix.
• src/registry.rs:24 bounds a log walk at MAX_LOG_HOPS, and no test reaches that branch. Worth a negative test later.