keep code as DRY as possible without changing public API signatures by only maintaining one function, `add_tags_to_tagset_builder`, that can add either one or many tags to a tagset.

the `add_one` and `add many` endpoints have the same behavior, but are kept as separate functions to prevent altering function signatures.

allow optionally getting canonical name when using `noit_metric_add_implicit_tags_to_tagset`.

add tests for implicit tags.

… realloc memory that was never allocated with malloc

… time

calloc instead of malloc.

free `stags` and `mtags` since they are initialized with calloc.

should not use `mtev_dyn_buffer_size` because the return is double the actul size we intent to use. use the value used to ensure the size instead.

should not use `mtev_dyn_buffer_used` when modifying the buffer through noit functions instead of buffer functions because return will be inacurate.

use `size_t` insted of `int32_t`.

use `mtev_dyn_buffer_data` instead of creating new pointer.

use const.

clean up tests and assert the expected tag pair sizes.

…tag`

…_completer` checks for `__check_uuid`.

properly initialize tagset

`tagset_implicit->tags` was being set through `MKTAGSETCOPY`. we do not need to copy the tags for `tagset_implicit` because it is an empty set and we are not using anything from `id` that could be changed.

use `calloc` if memory has not been allocated already.

.tag needs to be given it's own memory block so that it persists in the tagset after the input goes out of scope.

+1 to add a null terminator

the previous realloc was always for the same size that we just calloced.

This reverts commit 2787251.

This reverts commit 36f9db3.

…lass_t`.

add implicit tags to existing tagsets, but increase the max size to allow for them.

we can increase the max size because these tagsets are copied locally and the extended size does not conflict with the functions performed on the tagset in this scope.

since it is a copy, the implicit tags are not being added to the id's actual check and stream tags.

no need to get the encode(decode) length because they are inverse operations

… places

this fixes a segfault when attempting to realloc this memory later.
this also allows us to allocate only the necessary memory instead of the max.