chore: distinguish default/mtls endpoint for tests

[arithmetic1728]

showcase has default endpoint localhost:7469, so mtls endpoint is the same. When we interchange client.DEFAULT_ENDPOINT and client.DEFAULT_MTLS_ENDPOINT by mistake, the unit tests still pass. This causes problem, because generated client lib always has *.googleapis.com as the default endpoint, so mtls endpoint is always different (*.mtls.googleapis.com), so client lib unit test always fail. This PR replaces the default endpoint in case it is localhost (as it does in showcase unit tests) so we can distinguish the endpoints in showcase unit test.

[codecov]

Codecov Report

Merging #551 into master will not change coverage.
The diff coverage is n/a.

@@            Coverage Diff            @@
##            master      #551   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           26        26           
  Lines         1477      1508   +31     
  Branches       308       308           
=========================================
+ Hits          1477      1508   +31     

Impacted Files	Coverage Δ
gapic/schema/api.py	100.00% <0.00%> (ø)
gapic/schema/imp.py	100.00% <0.00%> (ø)
gapic/utils/cache.py	100.00% <0.00%> (ø)
gapic/schema/metadata.py	100.00% <0.00%> (ø)
gapic/schema/wrappers.py	100.00% <0.00%> (ø)
gapic/generator/options.py	100.00% <0.00%> (ø)
gapic/samplegen/samplegen.py	100.00% <0.00%> (ø)
gapic/samplegen_utils/types.py	100.00% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 71a7062...4a36e0e. Read the comment docs.