This changes `memalign` (and `posix_memalign`) so that it uses an
allocation method with the same selection criteria (checking
`malloc_alwaysinternal_limit` and picking one of:
- always MALLOC_CAP_INTERNAL
- MALLOC_CAP_INTERNAL first with fallback
- MALLOC_CAP_SPIRAM first with fallback
`malloc_alwaysinternal_limit` is in turn set by the options
`CONFIG_SPIRAM_MALLOC_ALWAYSINTERNAL` and
`CONFIG_SPRIAM_USE_CAPS_ALLOC`.
This notably affects folks using esp-rs to build rust code for the
esp-idf, as all allocations from rust use `memalign`.
Merges https://github.com/espressif/esp-idf/pull/12375
Users of the nvs API are likely to have `nvs_handle_t` in all cases, but
not all of them carry around the partition name and namespace name (as
they aren't needed after creating an `nvs_handle_t`).
Allow this common case to use nvs iteration without them tracking
additional data by using the `nvs_handle_t` to locate the namespace
index and the partition name by introducing an alterate to
`nvs_entry_find` called `nvs_entry_find_in_handle`, which operates
similarly except that it is given a `nvs_handle_t` instead of partition
and namespace strings.
This is somewhat more limited than the `nvs_entry_find` API as one
cannot examine all keys in a given partition.
With some libc impls (ie: Xcode 13.1), sprintf is marked as deprecated
and causes the build with `-Werror` to fail. Switch the `sprintf` usage
in test_nvs.cpp to use `snprintf` to allow builds to complete without
warning.
Cody P Schafer