add grow-case tests for squashfs and FAT32 copyFilesystems paths

[eriknordmark]

Summary

Two new tests covering the grow case (target partition larger than source) for the two filesystem types EVE-style partition migrations actually exercise:

TestCopyFilesystemsSquashfs

Builds a real squashfs into a source partition using go-diskfs's writer, then grows it 3× via copyFilesystems. Hits the squashfs branch, exercises the full raw-copy + verifyBlockCopy + squashfs.Read round-trip, with file content verification at the end.

Depends on diskfs/go-diskfs#404. Against the current upstream go-diskfs HEAD (7b44905), this test fails because:

• squashfs.Finalize writes the squashfs at byte 0 of the disk instead of the partition's byte offset (clobbering the GPT).
• sync.verifyBlockCopy would reject the grow case afterward because target size > source size.

With #404 applied (Sub-wrapping the backend in squashfs.Create/Read; LimitWriter on both sides in verifyBlockCopy), the test passes:

$ go test -run TestCopyFilesystemsSquashfs ./...
copying data from original partition 1 to new partition 2
partition 1 -> 2: performing raw data copy for filesystem type 2
partition 1 -> 2: contents copied byte for byte, 8388608 bytes copied
partition 1 -> 2: block copy verified
--- PASS: TestCopyFilesystemsSquashfs (0.12s)

TestCopyFilesystemsFat32Grow

Builds a FAT32 source with a known marker file and grows it 4×. Hits the FAT32 branch of copyFilesystems (CreateFilesystem + CopyFileSystem + CompareFS), which the existing TestCopyFilesystems didn't cover — that one uses an equal-sized target.

Does not depend on #404. Passes against unmodified upstream go-diskfs as-is; the FAT32 path uses file-level copy, not the raw-copy / verifyBlockCopy code that #404 touches.

Why this matters

Together, these two tests pin down the two filesystem-copy code paths that real grow workflows hit:

• squashfs rootfs partitions grown to make room for a larger replacement,
• ESP (FAT32) partitions grown to fit larger bootloader payloads.

Without them, copyFilesystems' grow behavior was only exercised end-to-end by TestRun, which is currently blocked on a separate go-diskfs ext4 multi-GB CreateFilesystem issue (diskfs/go-diskfs#402) and so doesn't surface regressions in the squashfs/FAT32 paths.

Merge order

This PR should land after diskfs/go-diskfs#404 merges and partitionresizer's go.mod bumps to a version that includes the fix. Until then, CI will report TestCopyFilesystemsSquashfs failing here.

The FAT32 test can be cherry-picked or split out earlier if you'd prefer; happy to follow either path.

Signed-off-by: eriknordmark erik@zededa.com