IFTSTA: path column not unique enough for stable diff JOIN key

[hf-mrdachner]

Context

In #256 / #257 we replace id_path with (path, segment_code) as the JOIN key in v_ahb_diff. This works for most message types but IFTSTA (21045) has structural path duplicates.

Problem

IFTSTA uses a flat segment group naming scheme (CNI-LOC-CNT-SG15) instead of the hierarchical paths used by other messages. This produces up to 5 duplicate (path, segment_code, qualifier) tuples per Prüfidentifikator — the STS segments under "Grund der Privilegierung nach EnFG" all have the same path.

Impact

• 36 IFTSTA-specific duplicates across all versions (out of 1676 total version/pruefi combinations)
• Non-IFTSTA duplicates exist but are benign (cnt=2 repetitions of identical elements like PARTIN FII "Name des Kontoinhabers" or MSCONS segment_group "Referenzangaben")

Non-IFTSTA duplicates (benign, cnt=2 only)

Nachricht	Prüfis	Segment	Detail
PARTIN	37000-37014	FII	Two "Name des Kontoinhabers" (D_3192) repetitions under Bankverbindung
MSCONS	13017	(segment_group)	Two "Referenzangaben" segment groups
PRICAT	27002	IMD	Two IMD segments under same path

These have identical content in both repetitions, so the diff result is still correct even with a false match.

Suggested fix

The IFTSTA path column needs to carry more structural context to be unique. This may need to be addressed in the materialization step where path is computed.

Related: #256, #257

[hf-kklein]

Gut beobachtet! Ich glaube wegen dieser Vielfalt an Daten ist es so schwer den id path ausnahmslos sauber zu konstruieren

[hf-kklein]

Addressed in #260. The IFTSTA case (flat SG naming, multiple STS segments with the same qualifier) is handled by the #N occurrence counter fallback. Since the qualifier-based approach can't distinguish these truly identical STS segments, the counter assigns #1, #2, etc. This is stable as long as the order of identical siblings doesn't change between versions.

The IFTSTA case is explicitly documented in the SQL comments and tested in the new test_id_path_uniqueness_per_message_type[IFTSTA-flat-SG-paths] test.

Great catch identifying this edge case, @hf-mrdachner!