This allows for appending a ':' and a positive integer to a requested worker type for the SYNAPSE_WORKER_TYPES
environment variable(and likewise, the WORKER_TYPES) used in Complement. For example:

'event_persister:2' would be yield identical results of 'event_persister, event_persister'

In the Synapse-worker docker image, multiple worker types are defined but only allowed to be that one single definition.
This makes it so that any two(or more) worker types can be merged into a single worker. For example:
1. If you wish to have one worker with both client_reader and federation_sender functions enabled, you can set
SYNAPSE_WORKER_TYPES='client_reader+federation_sender'
2. If you wish to have all stream_writers(excepting event_persister) you can set
SYNAPSE_WORKER_TYPES='account_data+presence+receipts+to_device+typing'

Multiple types can be combined, but some error checking to dis-allow multiples of worker types that shouldn't be
enabled more than once has been added. For example:
SYNAPSE_WORKER_TYPES='background_worker+event_persister, background_worker+event_persister' will not work as
background_worker is only allowed to have a single worker for the entire deployment.

Giving a worker or a combination of workers a custom name is as simple as adding the name then an equal sign in front
of the given worker type(s). For example:
SYNAPSE_WORKER_TYPES='alice=federation_reader'
or
SYNAPSE_WORKER_TYPES='bob=federation_inbound+federation_sender'
or
SYNAPSE_WORKER_TYPES='charlie=event_persister:2, derek=media_repository + pusher + user_dir + appservice + event_creator'

Instead of adding stream writer workers individually, utilize new functionality to make one worker with all included. Also utilitize new shorthand for event_persister's.

1. Turn 'placeholder_name' into a magic constant.
2. Fix casting on a worker_type_list from 'list' to 'List[str]'.
3. In 'insert_worker_name_for_shared_config', use new WORKER_PLACEHOLDER_NAME constant and reduce complexity by using value from items() directly.
4. In 'is_sharding_allowed_for_worker_type', simplify the return value.
5. Adjust 'split_string_and_strip' to allow for using the 'maxsplit' kwarg on 'strip' and update comment.
6. Remove function 'increment_counter' and use 'defaultdict' instead to reduce complexity.
7. Remove function 'is_name_allowed_for_worker', as this can be accomplished with an 'if' statement and only checked for a worker base name in a
    dict. Adjust 'generate_worker_files' to accommadate.
8. Rename a few dict's to remove misnomer's and adjust comments for extra clarity.
9. In 'generate_worker_files':
  a. Use 'ValueError' exception handling to catch additional errors when parsing an integer and simplify counting of appended worker_type's.
  b. Add additional error cases to check(single and double quotes, spaces, and last character being an integer) for potential file name/nginx issues.
  c. Adjust various variable names to improve readability.
10. Update various comments throughout.

1. Factor processing of all worker types from the environment out and away from
    generating config files.
2. Fix up early templating of 'worker_config' to remove boilerplate.
3. Wrangle nginx upstream processing to accommodate odd combinations for overlapping
    workers. e.g. 'user_dir, user_dir+presence' should be able to handle the(in this case
    single) endpoint over either worker. I think this was promised in a previous commit,
    consider it delivered.
4. Update a bunch of comments, and adjust some pre-existing to fit inside the line
    length count of 88(aka, make the green squiggles go away)
5. Move processing of unusable characters for worker names to a later position so it
    will check the name produced if one was not requested as well.

…rker_type', and similar and that should have been 'worker_types' when in use by sets and a couple other places.

1. Update comments and edit(or create) docstrings and a single type hint that was incomplete(split_and_strip_string()).
2. Rename parse_worker_types_from_env to parse_worker_types as it's not coming from the environment.

…, and update comments to match.

...so that even with a single port, a worker gets it's own upstream. Nginx says this can
lead to performance improvements if some features are enabled. Revisit this later.

1. Update a few comments for clarity.
2. Apply a replace() to remove whitespace in a requested worker name and use an underscore instead.
3. Stop-error if requesting a worker name, but forgetting to actually put one in.
4. If a worker name isn't requested, just use the set of worker types, sorted and concatenated with a hyphen. This avoids having to sanitize the worker name in this case.
5. Add regex to avoid nasty name surprises.

…it can't exist.

…ens if more than one worker type for a given worker is requested.

…hitespace on error message.