fix(auth): pass project id to gax clients

[callmehiphop]

Fixes #288
Fixes #445

• Tests and linter pass
• Code coverage does not decrease (if any source code was changed)
• Appropriate docs were updated (if necessary)

---

Looks like not passing the project ID into the gapic clients causes a bunch of future getProjectId calls which makes CPU usage spike pretty high

 [Summary]:
   ticks  total  nonlib   name
   3380   67.2%   71.3%  JavaScript

 [Bottom up (heavy) profile]:
   ticks parent  name
    922   18.3%  LazyCompile: *auth.getProjectId /Users/taco/projects/temp/node_modules/@google-cloud/pubsub/build/src/index.js:599:36
    921   99.9%    LazyCompile: *getProjectIdAsync.then.r /Users/taco/projects/temp/node_modules/google-auth-library/build/src/auth/googleauth.js:73:43
    921  100.0%      T node::RunMicrotasks(v8::FunctionCallbackInfo<v8::Value> const&)
    921  100.0%        LazyCompile: ~_tickCallback internal/process/next_tick.js:41:25

[stephenplusplus]

Do the GAPIC clients just need to cache the project ID that they fetch? In any case, this PR is correct-- we should provide to the GAPIC what the user provides, which could be different from what the environment detects. Just curious if there's a simple optimization that would make GAPIC perform better.

[callmehiphop]

@stephenplusplus The auth client should cache the project ID (at least it looks like it does). When a gapic client is created it makes an auth.getClient call for every service stub it creates. I haven't looked too deep into this, but it seems like we might be hitting the auth client with concurrent requests for credentials, which might explain the spike in CPU usage. If I had to guess, I think caching the auth client would help. Alternatively I think we could also pass our auth client to the gapic client, at which point the credentials would already be cached.