Fixing unit_test failure detection, and tests for data converters

[Groxx]

Apparently these also failed on the original PR, but due to missing a \ in #1303 it wasn't failing in CI because the env var was gone by the time exit ran.
I should really take that as a hint to finally rewrite that chunk of the makefile, so it just does a normal go test ./... :\ but not today.

Thankfully this seems to be the only set of tests that snuck past, and they were just incorrect.
I'm not sure why I apparently didn't run them or something while writing them in #1331 but it seems pretty clear I didn't since the old pre-merge SHA fails too. Bleh.

---

To correct the tests' original flaws:

1. memos are supposed to be encoded by custom dataconverters, and they are. we don't search them so they're not JSON like search attrs are, they're just blobs we expose in list APIs (and in workflow metadata).
2. Equal(string, []byte) just doesn't work, and the default dataconverter adds a trailing newline to separate args.
3. the []byte-heavy copypasta meant the Input-checking tests were odd, so I changed them, and fixed the args to the dataconverter. Internally we splat the args in here, so it's not []interface{..}-packed:

   cadence-client/internal/internal_worker.go

   Lines 573 to 579 in 6decfc7

   	// encode multiple arguments(arguments to a function).
   	func encodeArgs(dc DataConverter, args []interface{}) ([]byte, error) {
   	if dc == nil {
   	return getDefaultDataConverter().ToData(args...)
   	}
   	return dc.ToData(args...)
   	}

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 73.26%. Comparing base (f5ec359) to head (ba4e2ac).

Additional details and impacted files

see 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f5ec359...ba4e2ac. Read the comment docs.

[Groxx]

copied this "report the packages" from server, and ran it all with failing and fixed tests to double check :|
it's much nicer for reading when something fails, and "on the same line" like this is hopefully a bit more resilient to the bug here.

[Groxx]

previously I thought this was a general thing, but no! apparently not.

this might be something we can change in v2, since I don't think anyone really depends on it... or at least change the default. an explicit "no, encode byte slices too" seems fine, it's just unfortunate to lose the escape hatch without opting into it.