replace_packet_filters via async Gx CCR-U MODIFICATION (#22 Increment 5)

[next-nf]

Stacked on #54 (base async-ue-bearer-22-inc4). Fifth increment of #22.

Handles a UE replace_packet_filters Bearer Resource Command: report each replaced filter to the PCRF as a Gx CCR-Update MODIFICATION (naming the existing filter's SDF handle + carrying its new content), await the rule change, and emit a PTI-echoing Update Bearer Request with the changed TFT — replacing the true -> reject replace_packet_filters hit before.

A delete/add hybrid — almost all reuse

MODIFICATION is the correct (and only single-message) encoding: a CCR-U's Packet-Filter-Operation is message-level, so delete+add can't coexist in one message; the grouped AVP carries Packet-Filter-Identifier (WHICH filter) + Packet-Filter-Content (the NEW value) together.

• New (tiny): smf_tft:flow_info_to_pf_modify_group/2 = the Inc4 ADD content group plus a Packet-Filter-Identifier (the SDF handle, from inverting sdf_to_pf). With a diameter_Gx_SUITE wire round-trip confirming id + content coexist on the wire.
• ue_replace_filters_proc/4 — delete's reverse-lookup (pf_ids_to_sdf + zip), add's content, MODIFICATION op, remove+install fold, and the shared ue_delete_outcome branch (empty → Delete edge / non-empty → Update). Reuses emit_ue_update_bearer/await_pfcp_modify/detect_modified_bearers/ue_update_outcome/br_ok/br_err/ccr_result/fold_cca verbatim.

Tests

• pgw_SUITE 132/132 (131 prior + 1 new: one-filter bearer → replace with new content → one Update Bearer Request echoing the PTI carrying the changed, still-one-filter TFT; the delete/add regression stays green). smf_tft_SUITE (modify-group unit), diameter_Gx_SUITE 9/9 (MODIFICATION round-trip). Default compile green.
• The one design risk — whether a same-Charging-Rule-Name reinstall with changed Flow-Information surfaces a TFT delta through gx_events_to_pcc_ctx(install) so detect_modified_bearers fires — is empirically confirmed by the passing e2e (the Update Bearer only fires if the fold saw the change).

Note

Executed inline (authored directly, comprehensive tests, no separate implementer/reviewer subagent — unlike Inc2–4).

Out of scope

Requested-QoS/GBR change (Increment 6). A replace whose PCRF response binds a new {QCI,ARP} — out of scope; the branch acks gracefully.

[next-nf]

Superseded by the async-invoke restack. This increment was re-authored onto invoke(pipeline_async) and collapsed into #56 (delete/add/replace together) on the redone #48 foundation. Closing in favor of #56.