iso9660: expose rich metadata via FileInfo.Sys()

[luthermonson]

Refs #301. Sibling of #393 (ext4) — same pattern, applied to iso9660.

What this does

filesystem/iso9660/directoryEntry.Sys() previously returned *RockRidgeInfo when Rock Ridge extensions were present on the entry, otherwise nil. That was a partial answer to #301: callers could only see uid/gid/mode/symlink, only on Rock Ridge ISOs, and the plain ISO9660 metadata (extent location, volume sequence, hidden flag, etc.) was invisible.

This PR replaces RockRidgeInfo with a richer iso9660.StatT that Sys() always returns, carrying both the plain ISO9660 directory-entry fields and the Rock Ridge fields when present.

Fields

Source	Field
ISO9660 directory entry	ExtAttrSize uint8
	Location uint32 (extent LBA)
	VolumeSequence uint16
	IsHidden bool
	IsAssociated bool
	HasExtendedAttrs bool
	HasOwnerGroupPermissions bool
Rock Ridge PX record	UID uint32
	GID uint32
	NLink uint32
	Inode uint32
Rock Ridge SL record	LinkTarget string
(any RR record found)	RockRidge bool

RR fields are zero-valued / RockRidge=false on non-RR ISOs.

Breaking change — RockRidgeInfo removed

RockRidgeInfo was added recently (in #353) and exported, so removing it is a breaking change. In-repo callers in 4 test files are updated as part of this PR. External callers will need to switch:

// before
if rri, ok := info.Sys().(*iso9660.RockRidgeInfo); ok {
    fmt.Println(rri.Symlink, rri.UID)
}

// after
if st, ok := info.Sys().(*iso9660.StatT); ok && st.RockRidge {
    fmt.Println(st.LinkTarget, st.UID)
}

The alternative — keeping RockRidgeInfo alongside the new StatT — would have left two competing return types for Sys() and split the metadata model in two. Given that RockRidgeInfo is fresh (one release) and only covers a subset of what callers actually want, supersession seems cleaner than coexistence. Open to keeping it as a deprecated alias if reviewers prefer.

Tests

New TestSysStatT (plain ISO via testdata/9660.iso) and TestSysStatTRockRidge (RR ISO via testdata/rockridge.iso). Existing Rock Ridge tests (combined_xorriso_test.go, finalize_test.go, rockridge_xorriso_test.go) updated to assert on *StatT instead of *RockRidgeInfo.

[deitch]

This is fine and makes sense. I have no issue with deprecating RockRidgeInfo. Once again, go module structure makes life easier.

2 issues:

1. This is still marked as draft; is it ready?
2. My only request - and it is minor - is that the member func func (de *directoryEntry) statT() *StatT is in fileinfo.go rather than in directoryentry.go. I know, there is no "law" that requires all members to be in a single file, and we probably have places where there are so many member funcs that we had to split it, but I don't think that is the case here, so would rather have it in directoryentry.go with the struct and the other funcs. I don't care if we leave fileinfo.go with just the type StatT struct in it and nothing else, or move that into directoryentry.go.

[luthermonson]

I will address your feedback later tonight

[luthermonson]

Moved statT() into directoryentry.go next to Sys() in a212379. Left the StatT struct in fileinfo.go (the second of the two options you offered). Ready for another look.