server: add error presenter to make context cancellation semantic

feature: add submodule + middleware for feature flags

Refactor config loading and server initialization

As is, LoadConfig() does some things that are not strictly related to
the configuration, such as parsing command line arguments. This has led
to a proliferation of different ways to load the config based on various
needs and also prevents tools that need a config but are not services to
use custom command line arguments.

This commit aims to decouple config loading from everything else and
do nothing but loading the config files.

On a high level, this commit:
- renames server.NewServer() to server.New()
- moves config.Debug and config.Addr into the server package
- moves crypto.InitCrypto() call into server.New()
- moves command line parsing into server.New(), using passed-in values
  rather than os.Args

The only changes required for services would be changing

   cfg := config.LoadConfig(":5100")
   server := server.NewServer("meta.sr.ht", cfg)

to

   cfg := config.LoadConfig()
   server := server.New("meta.sr.ht", ":5100", cfg, os.Args)

All other tools will be switched to just LoadConfig() and, optionally, a
call to crypto.InitCrypto(). I managed to completely remove some global
state (addr) and at least make the rest private, so that users are
forced to use the designated functions.

The config module gained support for custom FS implementation, mainly
for testing.

server: enable GraphQL introspection

server: log GraphQL queries if running in debug mode

server: handle CORS in server

Currently CORS is handled in nginx in front of the API. That setup is
not very suitable for Kubernetes. Instead, handle CORS here, so we do
not need any intermediaries between the ingress and the APIs.

The CORS settings are taken from sr.ht-nginx/graphql.conf [1]

[1] https://git.sr.ht/~sircmpwn/sr.ht-nginx/tree/master/item/graphql.conf

server: replace deprecated gqlgen code

Most of the gqlgen/handler package, specifically also
`handler.GraphQL()`, has been deprecated [1]. The documentation of the
new package [2] is a bit lacking, but some stuff is covered in the
reference docs [3] [4], and some in the examples [5].

[1] https://pkg.go.dev/github.com/99designs/gqlgen@v0.17.36/handler#GraphQL
[2] https://pkg.go.dev/github.com/99designs/gqlgen@v0.17.36/graphql/handler
[3] https://gqlgen.com/reference/errors/#the-panic-handler
[4] https://gqlgen.com/reference/complexity/#limiting-query-complexity
[5] https://github.com/99designs/gqlgen/blob/master/_examples/fileupload/server/server.go

server: disable logging for release mode

This was partially implemented a while ago but we missed the main point
where it matters.

server: add pprof endpoint

This adds a pprof endpoint to all SourceHut APIs. It picks an
OS-assigned port on localhost to run it on and prints this port to
stdout. This ensures a unique port is selected even if several APIs are
running on a single host.

server: enable request logging only in debug mode

To improve the signal:noise ratio in the production logs.

redis: avoid building UniversalOptions for standard mode

This makes it support all options that go-redis recognizes in ParseURL,
by constructing a standard client directly instead of building
UniversalOptions ourselves.

server/email: print correct var on type assertion failure

Export webhooks public key via api-meta.json

server: add @admin directive implementation

This will be used at least by meta.sr.ht in the near future.

config: factorize integer parsing

Add a new GetInt() function to parse an integer value from a parsed
ini.File. Use that function instead of duplicated code.

Signed-off-by: Robin Jarry <robin@jarry.cc>

treewide: update to new multi worker dowork api

The work.Queue() implementation has changed and now allows scheduling
tasks from multiple goroutines in parallel. Also, it requires a new
argument for limiting the queue buffer size in order to apply back
pressure.

Define new configuration variables to allow tuning the queue sizes and
number of workers per service:

  [mail]
  # Maximum size of the outgoing email queue (default 512).
  egress-queue-size = 512

  [$service_name]
  # Number of parallel workers per queue (default 1).
  # There are multiple queues (for egress email, webhooks, etc.).
  # This setting is applied on a per-queue basis.
  queue-workers = 1

  [webhooks]
  # Maximum size of the webhooks queue (default 512).
  queue-size = 512

Use these new settings to configure the queues and workers accordingly.

Fix unit tests as well.

Link: https://git.sr.ht/~sircmpwn/dowork/commit/95719cfc0118
Signed-off-by: Robin Jarry <robin@jarry.cc>

redis: use UniveralClient, support sentinel mode

This commit switches the redis client to using a UniversalClient [1],
which can operate in standard, cluster, or failover (sentinel) mode.

It also implements basic support for having (multiple)
`redis+sentinel://` URLs as connection URL in the config (e.g. for
`redis-host`).

This is the Go equivalent of the proposed Python patch [2]. It supports
the same examples given there.

[1] https://pkg.go.dev/github.com/go-redis/redis/v8#UniversalClient
[2] https://lists.sr.ht/~sircmpwn/sr.ht-dev/patches/55521

server: add AnonRouter function

To expose anonymous (unauthenticated) routing to downstream core-go
users.

server: fix playground on config.Debug

A recent change made it so that the playground required authentication
to access in debug mode; this moves it under the other auth-free
endpoints.

server: use routing groups

Instead of hardcoding some exceptions in the auth middleware, use a
different routing group for routes that do not require auth. Makes the
auth middleware more generic and also removes a lot of unneccessary
middleware processing from routes that don't need it.

For now, the added group is not accessible from outside the module, but
if the need arises, this might be an option.