fix(cdd): rebuild CDD import on cddtools, read field values correctly (#422)

[crazy0104]

Fixes the CDD half of #422.

The bug

Importing a tester from a CDD died with could not convert string to float: '0,0,0,0,0,0'.

A data type can be a field (array): CVALUETYPE@bl is then the width of one element, minsz/maxsz the element count ("Bit Length / Field Size" in the CANdelaStudio help). A value of such a type lists one entry per element, in CANdela's list notation (0,0,0,0,0,0) — the same notation the document uses for languages='(en-US,zh-HANS)' or mayBeExec='(1,2,3,4,5)'. So any attribute value can be a list, and reading one as a single number aborted the whole document.

The old parser hit it in _load_choices (a text table over a field type). It was not the only unguarded spot, which is why this replaces the numeric handling wholesale rather than patching one call site.

What changed

resources/cdd/cdd_tester_parser.py, cdd_parser_base.py and cdd_shared.py (~2600 lines, the second and third being near-duplicates of each other) are replaced by the vendored cddtools package under resources/cdd/cddtools/. cddparse.py is now a thin CLI shim. The package is pure stdlib, so nothing changes for the build; provenance and re-sync steps are in resources/cdd/cddtools/README.md.

Every CDD attribute value now goes through one decoder that returns either a scalar or the element list, folding elements big-endian by element width, and never lets a raw string reach int()/float().

Behaviour

Same testers, same service ids, same service count and names, same CAN addresses and UDS timing as before on the same document. Deliberate differences:

• a payload container becomes its own parameters instead of one opaque byte array (SessionParameterRecord ARRAY 32 → P2 NUM 16 + P2Ex NUM 16)
• a DID read expands into one parameter per DID member, each keeping its own name (the old parser emitted only the first member)
• components the document declares but the old parser dropped now appear (PowerDownTime on 0x11, Security Access Data Record on 0x27)
• two old defects fixed on the way: SecuritySeed was 16 bits where the data type says 32, and 0x28 EnableRxAndEnableTx lost its 0x00 subfunction because it is marked spec="accm", not spec="sub"

Tests

test/cdd/fixtures/CddTools-doip.cdd replaces the two hand-written fixtures — neither was a document CANdelaStudio could write. It is a template-derived document saved out of CANdelaStudio with CAN and DoIP both declared as supported, which covers what a synthetic file cannot: the variant x interface split, DoIP addressing, 96 services over 20 SIDs, subfunction locking, DID member expansion and an ASCII field parameter.

npm run test -- test/cdd 8/8, npm run typecheck clean.

Not in this PR

The PDX/ODX half of #422 (invalid literal for int() with base 16: '0x') is untouched: odxtools reads A_BYTEFIELD two hex characters at a time, so a value written 0x1A2B makes the first pair 0x. That needs a separate patch in resources/odx/odxparse.py.

[frankie-zeng]

Not in this PR
The PDX/ODX half of #422 (invalid literal for int() with base 16: '0x') is untouched: odxtools reads A_BYTEFIELD two hex characters at a time, so a value written 0x1A2B makes the first pair 0x. That needs a separate patch in resources/odx/odxparse.py. 这个是不是再来一个PR把他修复了

[crazy0104]

Not in this PR The PDX/ODX half of #422 (invalid literal for int() with base 16: '0x') is untouched: odxtools reads A_BYTEFIELD two hex characters at a time, so a value written 0x1A2B makes the first pair 0x. That needs a separate patch in resources/odx/odxparse.py. 这个是不是再来一个PR把他修复了

我先看看还有没有其他ODX/PDX解析问题，下一个PR修复。