feat(mp4): add WaveBox.GetChildren and escape control chars in box types

[tobbee]

Problem

Follow-up to #547, which added the wave box. Three small gaps were left:

• WaveBox holds Children but has no GetChildren, so unlike the other
  container boxes it does not satisfy the ContainerBox interface.
• The QuickTime terminator atom inside a wave has the four-byte type
  0x00000000, which Info() wrote out raw. Every mp4ff-info dump of a
  QuickTime audio file therefore contained NUL bytes and rendered the atom as
  [ ].
• The RawTail semantics were undocumented: a malformed atom anywhere in the
  payload puts everything from it onwards in the tail, so boxes after it are
  not decoded into Children.

Fix

• WaveBox.GetChildren, so a wave box is a ContainerBox. Its doc comment
  notes that RawTail is not part of the children, so a generic container
  encode would drop those bytes.
• displayBoxType in the info dumper escapes C0 control characters and DEL as
  \xNN. It deliberately leaves 0xa9 alone, so the ©nam-style Apple atoms
  keep rendering as before and no golden dump changes. While there,
  fixStartingCopyrightChar is guarded against an empty box type, which would
  panic on the index of byte 0.
• RawTail is documented, including that Encode writes it verbatim after
  the children.

mp4ff-info on an ffmpeg-written .mov, before and after:

-                [    ] size=8
+                [\x00\x00\x00\x00] size=8
                  - not implemented or unknown box

Tests

TestWaveBoxIsContainer asserts the ContainerBox interface and the
GetChildren contents; TestWaveBoxTerminatorInInfo asserts that no raw NUL
reaches the output and that the escaped form appears. go test ./...,
go vet, gofmt, and golangci-lint pass.

Also checked outside the test suite: an ffmpeg-generated AAC-in-.mov file
(which carries a real wave atom with frma, a 12-byte mp4a stub, esds,
and the terminator) still round trips byte-exact through both the
io.Reader/io.Writer and the SliceReader/SliceWriter paths.

🤖 Generated with Claude Code