fs.stat: a symlink's size is its stored object length (target + 27 bytes); is the target length intended? #447

Summary

fs.stat reports a symlink's size as the length of the canonical object that stores its target, not as the target's length. That object holds the target plus the object envelope (LFSO, 13 bytes) and the symlink header (LFS4LNK, 14 bytes), so every symlink comes out 27 bytes too large. A mounted Workspace reports the target's length, and so does the source folder Init read from, which means the SDK and the mount disagree about the same symlink.

Repro

Run on Linux against current main (c77281e49) with the real addon and a Disposable Store. Init a folder holding link -> a.txt and dangling -> /does/not/exist, then call:

fs.stat("main", "/link")      -> { type: "symlink", size: 32 }   // expected 5
fs.stat("main", "/dangling")  -> { type: "symlink", size: 42 }   // expected 15

ls -l in the source folder shows 5 and 15. Under a real FUSE mount (the kernel_mount fixture's five symlinks), lstat size equals the readlink length for each one.

Cause

For InodeKind::Symlink, size() in layerfs-store-service/src/read/snapshot.rs returns read_canonical(content_root).len(). The Workspace base view does it differently: stat_resolved_with_lengths in layerfs-workspace/src/base/view.rs decodes the object with SymlinkTarget::decode and uses as_bytes().len(), and the FUSE attributes take that value. The operator file history and stats also call size(), so they count symlinks 27 bytes too large as well.

Why I think the target length is intended

• The doc comment on size() says "Logical size of a file or symlink's content". The regular-file branch has returned the logical length since SDK redesign (JS-only over a Rust implementation by group/method), commit-time merge and conflicts, and live wiring of the sdk-sim-terminals UI #327 (7d), but a symlink falls into the _ => branch, which returns canonical.len().
• The SDK redesign (JS-only over a Rust implementation by group/method), commit-time merge and conflicts, and live wiring of the sdk-sim-terminals UI #327 plan defines canonical bytes as including framing. It reports them under their own names (canonical, newCanonical), separate from size.
• The Workspace view calls the same value logical_len, and for a symlink sets it to the target length.

If the canonical length is what fs.stat should report, I can change the PR to document that instead.

Expected

A symlink's fs.stat size is its target's byte length, the same as the mount reports.

Fix with a regression test: #448.