Fix CTZ traverse pointer conversion count

[Old-Ding]

Summary

• convert only the CTZ head pointers that lfs_bd_read actually filled
• move the conversion after the read error check, so failed reads do not decode local buffer contents
• avoid reading an uninitialized heads[1] slot when count == 1

Fixes #1214

Testing

• git diff --check
• git ls-files --eol lfs.c
• Not run: make test requires the project test toolchain, but this Windows host does not have make, cc, or gcc available.

[geky-bot]

Tests passed ✓, Code: 17200 B (+0.0%), Stack: 1448 B (+0.0%), Structs: 812 B (+0.0%)

	Code	Stack	Structs		Coverage
Default	17200 B (+0.0%)	1448 B (+0.0%)	812 B (+0.0%)	Lines	2448/2609 lines (-0.0%)
Readonly	6234 B (+0.0%)	448 B (+0.0%)	812 B (+0.0%)	Branches	1301/1638 branches (-0.0%)
Threadsafe	18056 B (+0.0%)	1448 B (+0.0%)	820 B (+0.0%)		Benchmarks
Multiversion	17272 B (+0.0%)	1448 B (+0.0%)	816 B (+0.0%)	Readed	29000746676 B (+0.0%)
Migrate	18864 B (+0.0%)	1752 B (+0.0%)	816 B (+0.0%)	Proged	1482895246 B (+0.0%)
Error-asserts	18036 B (+0.0%)	1440 B (+0.0%)	812 B (+0.0%)	Erased	1568921600 B (+0.0%)

[Kishore-Gorijavolu]

Reviewed the CTZ traversal change and independently tested it against both the pre-fix and PR revisions.

The change looks correct to me. In the existing implementation, count can be either 1 or 2, but both heads[0] and heads[1] were unconditionally passed through lfs_fromle32(). When count == 1, only one entry is populated by lfs_bd_read(), so converting heads[1] unnecessarily accesses an uninitialized element.

This patch addresses that cleanly by:

• checking the lfs_bd_read() result before consuming the returned data;
• converting only count entries using the bounded loop; and
• preserving the subsequent heads[count-1] traversal behavior.

I also performed a focused before/after test using the same test source against both revisions. The test created a 4096-byte multi-block file and exercised lfs_fs_traverse(). Both revisions completed successfully with the same 19-block traversal sequence, indicating no functional regression in the exercised CTZ traversal path.

I attempted to run the complete test suite locally as an additional regression check, but the local Windows/MSYS Python environment could not execute the littlefs test infrastructure because the POSIX termios module was unavailable. I therefore did not treat that attempt as a project test result. The PR's GitHub CI checks are passing.

The implementation is appropriately scoped to the pointer-conversion and read-error handling path.

Approved.

PR-1217.zip