Add support for version 1 (0x501) keytab files (#5)

[p0dalirius]

Linked Issue

Closes #5

Root Cause

The keytab (de)serialization was written for version 2 only. Every integer read and write used binary.BigEndian, the name_type 32-bit field was always present, and the component count was used verbatim. Version 1 of the format differs in exactly three ways — it uses native (little-endian) byte order, its component count includes the realm, and it has no name_type field — and none of these were derived from FileFormatVersion. As a result version-1 files were parsed with the wrong byte order and a phantom name_type, desynchronizing every field, and the writer could only emit version-2 layout.

Fix Description

• Added src/keytab/FileFormatVersion.go with the version constants, a byteOrderForVersion helper (version 1 → little-endian, otherwise big-endian), and small resolvers for the optional parameters.
• CountedOctetString and KeyBlock FromBytes/ToBytes take an optional binary.ByteOrder and use it for all integer fields.
• KeytabEntry.FromBytes/ToBytes/UpdateSize take an optional version; they derive the byte order, subtract 1 from the on-disk component count for version 1 (and add it back when writing), and skip the name_type field for version 1, defaulting it in memory to KRB5_NT_PRINCIPAL.
• Keytab.FromBytes/ToBytes/UpdateEntriesSizes pass FileFormatVersion down to the entries. The 2-byte version header itself is unchanged (always 0x05 followed by the version byte).

All new parameters are variadic and default to version 2 / big-endian, so the change is backward compatible: existing call sites (and all existing tests) behave exactly as before.

Note on byte order: the spec says version 1 uses "native" byte order. Since a reader cannot know the producing host's endianness, this implementation assumes little-endian, which is the conventional interpretation (version 1 predates the big-endian convention that version 2 introduced to remove the ambiguity).

How Verified

• go test ./... and go vet ./keytab/ pass.
• keytab describe -f example.kt (version 2) is unchanged.
• New tests parse a hand-built little-endian version-1 record and confirm the byte order, component-count adjustment, and absent name_type are all handled, then confirm byte-exact round-trips.

Test Coverage

• Added: src/keytab/Keytab_version1_test.go:
  • Test_KeytabEntry_Version1_Parse — a hand-built little-endian record decodes correctly (realm subtracted from the count, name_type defaulted, all fields little-endian).
  • Test_KeytabEntry_Version1_RoundTrip — that record serializes back to identical bytes.
  • Test_Keytab_Version1_FileRoundTrip — a full version-1 keytab round-trips and emits the 0x05 0x01 header with a little-endian, realm-inclusive component count.
  • Test_KeytabEntry_Version1_OmitsNameType — a version-1 entry is exactly 4 bytes shorter than the version-2 encoding of the same entry.

Scope of Change

• Files changed: src/keytab/FileFormatVersion.go (new), src/keytab/CountedOctetString.go, src/keytab/KeyBlock.go, src/keytab/KeytabEntry.go, src/keytab/Keytab.go, src/keytab/Keytab_version1_test.go (new)
• Submodule pointer updated: no
• Behavioral changes outside the bug fix: none — version-2 behavior is byte-for-byte identical.

Risk and Rollout

The optional-parameter defaults preserve all version-2 behavior, so the blast radius is limited to the new version-1 path. Safe to merge directly.

Notes

This branch is based on main and touches the same (de)serialization methods as the open PRs for #1 (parser robustness) and #2 (kvno round-trip); whichever merges first, the others will need a straightforward rebase.