* Sync: Bump indexer to recognize pqsig signatures

Nightly algod now produces transactions signed with post-quantum
signatures. idb.SignatureType in the pinned indexer only knew
sig/msig/lsig, so the postgresql exporter aborted the round:

  AddBlock() err: getSigTypeDelta() err: unable to determine the
  signature type for stxn=&{... PQsig:{Scheme:[102 49] ...}}

Indexer main handles this as of 219b2118 ("Update to recognize
pqsigs"). It pins the same go-algorand-sdk revision we already use, so
this bump touches nothing else.

* Tests: Stop leaking a running pipeline out of TestHealthEndpoint

The test started the pipeline in a goroutine and never stopped it, so it
kept writing metadata into the data dir while t.TempDir() cleanup was
removing it:

  --- FAIL: TestHealthEndpoint/API_OFF
      TempDir RemoveAll cleanup: unlinkat /tmp/...: directory not empty

There was no way to stop it: runConduitCmdWithConfig hardcoded
context.Background(). Take a context instead, which MakePipeline already
derives its cancelable context from, and have the test cancel and wait
for the goroutine to finish before returning. The cobra command passes
cmd.Context(), which cobra defaults to context.Background(), so the
binary behaves as before.

* comment typo

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

The test started the pipeline in a goroutine and never stopped it, so it
kept writing metadata into the data dir while t.TempDir() cleanup was
removing it:

  --- FAIL: TestHealthEndpoint/API_OFF
      TempDir RemoveAll cleanup: unlinkat /tmp/...: directory not empty

There was no way to stop it: runConduitCmdWithConfig hardcoded
context.Background(). Take a context instead, which MakePipeline already
derives its cancelable context from, and have the test cancel and wait
for the goroutine to finish before returning. The cobra command passes
cmd.Context(), which cobra defaults to context.Background(), so the
binary behaves as before.

Nightly algod now produces transactions signed with post-quantum
signatures. idb.SignatureType in the pinned indexer only knew
sig/msig/lsig, so the postgresql exporter aborted the round:

  AddBlock() err: getSigTypeDelta() err: unable to determine the
  signature type for stxn=&{... PQsig:{Scheme:[102 49] ...}}

Indexer main handles this as of 219b2118 ("Update to recognize
pqsigs"). It pins the same go-algorand-sdk revision we already use, so
this bump touches nothing else.

The PR workflow pins CI_E2E_FILENAME to a commit hash, so its block data
is frozen and cannot surface encoding changes in go-algorand. New block or
transaction fields are fatal rather than ignorable, because the SDK decodes
with ErrorIfNoField set, so an unknown map key aborts block import. The
unit tests cannot catch it either: they encode and decode with the same SDK
structs, which round-trips regardless of what algod puts on the wire.

The PR workflow pins CI_E2E_FILENAME to a commit hash, so its block data
is frozen and cannot surface encoding changes in go-algorand. New block or
transaction fields are fatal rather than ignorable, because the SDK decodes
with ErrorIfNoField set, so an unknown map key aborts block import. The
unit tests cannot catch it either: they encode and decode with the same SDK
structs, which round-trips regardless of what algod puts on the wire.

The nightly algod now emits the pqsig key in SignedTxn. The previously
pinned SDK (2026-06-24) predates the post-quantum signature types, so
go-codec rejected the unknown key and block import failed with
"no matching struct field found when decoding stream map with key pqsig".

The nightly algod now emits the pqsig key in SignedTxn. The previously
pinned SDK (2026-06-24) predates the post-quantum signature types, so
go-codec rejected the unknown key and block import failed with
"no matching struct field found when decoding stream map with key pqsig".

* Don't allow filtering on the pq signatures

recursiveTagFields passed the raw codec tag down when recursing into a
struct field, so a tag like `codec:"txn,required"` produced tag paths of
the form `txn,required.*`. Every ignoreTags lookup for a nested field
then missed, and the generated filter tag would carry the encoder option
in its name.

Parse the tag name once, before recursing, and use it for both the
nested path and the leaf's full tag. A tag that only carries options
(`codec:",omitempty"`) names nothing, so it is now treated as untagged
at any level rather than only at leaves.

No SDK struct-typed field carries tag options today, so the generated
map and tags doc are unchanged.

recursiveTagFields passed the raw codec tag down when recursing into a
struct field, so a tag like `codec:"txn,required"` produced tag paths of
the form `txn,required.*`. Every ignoreTags lookup for a nested field
then missed, and the generated filter tag would carry the encoder option
in its name.

Parse the tag name once, before recursing, and use it for both the
nested path and the leaf's full tag. A tag that only carries options
(`codec:",omitempty"`) names nothing, so it is now treated as untagged
at any level rather than only at leaves.

No SDK struct-typed field carries tag options today, so the generated
map and tags doc are unchanged.