- Replace TODO authoring notes with user-facing summaries (or drop
  entries that are not user-visible).
- Drop changelog item for `#101634` since the PR only adds a test.
- Drop changelog item for `#100315` since the PR only adds an
  internal `getLastWrittenObjectPath` method on `ObjectStorage`.
- Rewrite the duplicated "This PR addresses several issues..." entry
  for `#99163` / `#100420` as a single coherent sentence.
- Fix the `chdig` link target: the text said `v26.4.3` but the URL
  pointed to the `v26.4.1` tag. Both versions exist; the PR title is
  "Bump chdig to v26.4.3", so align the URL with the version.

PR: #103729

…solution in next commit)

---
Original cherry-pick message follows:

Merge pull request ClickHouse#100420 from ClickHouse/divanik/rerevert_spark_azure_fixes

Resolve problems with paths and compatibility problems with Spark in Azure (v2)

# Conflicts:
#	src/Interpreters/IcebergMetadataLog.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/IcebergMetadata.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/IcebergWrites.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/ManifestFileIterator.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/MultipleFileWriter.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/MultipleFileWriter.h
#	src/Storages/ObjectStorage/DataLakes/Iceberg/Mutations.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/PersistentTableComponents.h
#	src/Storages/ObjectStorage/DataLakes/Iceberg/Utils.cpp
#	src/Storages/ObjectStorage/DataLakes/Iceberg/Utils.h

Adapted PR ClickHouse#90740 (Read iceberg from various paths) to antalya-26.3
without applying the prerequisite upstream PR ClickHouse#100420 (IcebergPath /
path_resolver refactor). The refactor is dropped; raw `String` paths are
used instead.

Adaptations from PR 90740 to antalya-26.3:

- `IcebergPathFromMetadata` references → plain `String` (no `.serialize()`,
  no `IcebergPathFromMetadata::deserialize` wrapping).
- `IcebergPathResolver & path_resolver` parameters → `const String &
  table_location`. Calls like `path_resolver.resolve(x)` become `x`.
- `SecondaryStorages` infrastructure kept: thread-safe map of secondary
  object storages plus a `resolveObjectStorageForPath` helper that maps
  a metadata path to a (storage, key) pair. The IcebergPath-aware
  overload of `resolveObjectStorageForPath` was removed.
- New protocol version `DBMS_CLUSTER_PROCESSING_PROTOCOL_VERSION_WITH_ICEBERG_ABSOLUTE_PATH = 7`
  used in `IcebergObjectSerializableInfo::{serializeForClusterFunctionProtocol,
  deserializeForClusterFunctionProtocol}` to gate the new
  `data_object_file_metadata_path` field and `requires_external_storage`
  check; `_path` for delete files goes through `SchemeAuthorityKey` on
  older protocols.

Dropped (depend on upstream commits not on antalya-26.3):

- `ExpireSnapshotsExecute.{cpp,h}`, `RemoveOrphanFilesExecute.{cpp,h}`,
  `SnapshotFilesTraversal.{cpp,h}` — extracted EXECUTE handlers from
  upstream PR introducing per-command refactor. PR 90740 only threads
  `secondary_storages` into these; the underlying refactor is a
  separate dependency. The antalya-26.3 `Iceberg::expireSnapshots` path
  is kept unchanged in `IcebergMetadata::executeCommand`.
- `executeExpireSnapshots` / `executeRemoveOrphanFiles` dispatch in
  `IcebergMetadata::executeCommand` — depends on the dropped files.

References:

- Upstream PR: ClickHouse#90740
- antalya-26.1 backport (used as a structural reference for the
  no-IcebergPath adaptation): 0520e2e
  ("Allow to read iceberg table data from any location")

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The path resolution refactor in ClickHouse#100420/ClickHouse#100295 changed `FileNamesGenerator`'s
constructor signature (5 args → 4 args, with `is_transactional` mapped to
`use_uuid_in_metadata` and `config_path` removed) and removed
`generateMetadataName()` in favour of `generateMetadataPathWithInfo()`.
The `generateManifestList` signature also changed: first arg is now
`IcebergPathResolver` instead of `FileNamesGenerator`, and manifest entry
sizes is `std::vector<Int64>` instead of a scalar.

Adapt the `TRUNCATE TABLE` code path (from #1655) to the new API:

- Two-branch `FileNamesGenerator` construction (transactional vs. non-transactional,
  reading location from JSON) → single construction using
  `persistent_components.path_resolver.getTableLocation()` with
  `is_transactional` as `use_uuid_in_metadata`.
- `generateMetadataName()` → `generateMetadataPathWithInfo()`, returning
  `GeneratedMetadataFileWithInfo { path, version, compression_method }`.
- Storage paths derived via `path_resolver.resolve(path)` where raw strings
  were previously returned.
- Catalog filename via `path_resolver.resolveForCatalog(path)`, replacing the
  manual `location + metadata_name` concatenation.
- `generateManifestList(filename_generator, ...)` → `generateManifestList(path_resolver, ...)`.

Behavior of truncate is preserved: clears data files by writing a new
snapshot with empty manifests and updates the catalog pointer. The new
`IcebergPathResolver` abstraction handles both transactional (full URI)
and non-transactional (bare path) cases transparently.

Refs: #1785, #1655

…100420

Backport of ClickHouse#100420/ClickHouse#100295 - Resolve problems with paths and compatibility problems with Spark in Azure (v2)

…_spark_azure_fixes

Resolve problems with paths and compatibility problems with Spark in Azure (v2)