Kamu

Added
- Externally configurable Argon2 hashing mode (minimal settings to speedup e2e tests)
Changed
- Unified dataset repository format:
- both LFS and S3, regardless of tenancy config, now organize storage folders solely by dataset ID
- workspace migration to new repository format (v6) is fully automatic
- the role of "info/alias" file was reduced only for recovery purposes,
the dataset alias resolutions now happen in database only
- ODF storage unit implementations (LFS, S3) only deal with dataset identifiers:
- no more dependency on accounts or tenancy configuration
- S3 cache now considers identifiers of stored datasets only
- moved them to ODF crate
- reading HEAD is a must for storage iteration
- removing HEAD is a first step of dataset deletion
- `kamu-datasets` domain now has own error and result structures layer, separate from ODF basics
- Rename operation does not touch a storage unit, became solely a database operation
- workspace version checking now takes startup job dependencies into account
Fixed
- Less linear search in in-mem entry repository

Fixed
- E2E: `repo-tests` crate again contains the `kamu-cli` dependency, but as an optional one
- This way we can correctly reuse the tests in the `kamu-node` repository without affecting the build time
- It also fixes `make sqlx-prepate` developer command

Added
- Added prometheus metrics for AWS SDK S3 calls
Changed
- E2E: make `repo-tests` crate independent from `kamu-cli` crate

Fixed
- Fixed arrow `BinaryView` incorrectly being treated as incompatible with `Binary` fields during dataset schema compatibility checks (1096)

Added
- Added [common-macros](src/utils/common-macros) crate containing macros of general use
- `kamu list`: display dataset visibility in multi-tenant
- `kamu pull`: added `--visibility private|public` argument to specify the created dataset visibility
Changed
- Improved/added trace for repositories & GQL to contain not only the method name but also the structure name
Fixed
- Restoring OData API tolerance to trailing slashes
- OData API: fixed crash when accessing private dataset

Added
- The usage of local database (SQLite) is activated by default for all single tenant workspaces
Fixed
- Improved error message for SQL parsing method for queries which includes invalid or reserved keywords
- Fixed false-positive panic ("There cannot be predefined users in a single-tenant workspace")
if a `kamu` subcommand that doesn't require workspace found a multiuser `.kamuconfig`
- For example, before attempting to initialize a workspace or attempting to invoke autocomplete in shell