RCORE-2233 Reduce locking for StringInterner lookup and compare methods

[finnschiermer]

What, How & Why?

Reduce locking in the string interner:

• when using it for comparing strings or for mapping strings to StringIDs.
• when mapping StringIDs to strings.

Also entirely remove locking from methods where it isn't needed (less important since they are much less frequently used than the methods mentioned above).

[coveralls-official]

Pull Request Test Coverage Report for Build finn.schiermer-andersen_96

Details

• 45 of 45 (100.0%) changed or added relevant lines in 2 files are covered.
• 77 unchanged lines in 16 files lost coverage.
• Overall coverage increased (+0.005%) to 91.074%

---

Files with Coverage Reduction	New Missed Lines	%
src/realm/node_header.hpp	1	63.35%
src/realm/sync/network/http.hpp	1	82.27%
src/realm/sync/network/websocket.cpp	1	72.2%
test/fuzz_tester.hpp	1	57.73%
test/test_index_string.cpp	1	93.33%
test/test_query2.cpp	1	98.75%
src/realm/object-store/shared_realm.cpp	2	91.89%
test/test_lang_bind_helper.cpp	2	93.08%
src/realm/query_engine.cpp	3	92.1%
src/realm/bplustree.cpp	7	72.93%

Totals
Change from base Build jorgen.edelbo_375:	0.005%
Covered Lines:	220199
Relevant Lines:	241781

---

💛 - Coveralls

[jedelbo]

Would it make sense to have a counter that is increased/decreased when calling the non-threadsafe operation. Then we could have assertions requiring that the counter is 1 inside those functions and 0 in the threadsafe functions below.

[finnschiermer]

for such a counter to be correctly shared among threads (and thus be useful for asserting) it would have to be an atomic which carries some overhead. I'm not convinced it is worth it?

[finnschiermer]

I could add it in debug mode only?

[nicola-cab]

@finnschiermer curios to see how the benchmarks look like :-) .. I can run those.