Make coverage perfoms a check that the computed coverage is at least

not lower than the reference coverage, ensuring no regression
on the coverage of the tests can occur.

NOTE: some tests (m-i-shared) perform some dynamic test cases
where the exact coverage depend on the exact dynamic behavior
of the test. And this behavior is not always reproducible
(for exact the calls to m_core_backoff_wait may not occur
if the threads are not trying to grab the resource at the
exact same time).
As such, the reference of such tests is taken from the worst
case of execution of the coverage (with the lowest number
of lines covered).

tests/synthesis.ref

@@ -0,0 +1,27 @@
+|	          M-ALGO	|	 100.00% of 1160	|
+|	         M-ARRAY	|	   98.24% of 908	|
+|	        M-BITSET	|	   98.31% of 413	|
+|	       M-BBPTREE	|	   99.60% of 998	|
+|	        M-BUFFER	|	   98.86% of 877	|
+|	    M-CONCURRENT	|	  100.00% of 102	|
+|	          M-CORE	|	  100.00% of 268	|
+|	         M-DEQUE	|	   99.47% of 943	|
+|	          M-DICT	|	  93.00% of 1528	|
+|	       M-FUNCOBJ	|	  100.00% of 108	|
+|	        M-GENINT	|	    97.40% of 77	|
+|	        M-I-LIST	|	   98.10% of 474	|
+|	      M-I-SHARED	|	   100.00% of 112	|
+|	          M-LIST	|	  99.12% of 1368	|
+|	       M-MEMPOOL	|	    97.98% of 99	|
+|	         M-MUTEX	|	   100.00% of 19	|
+|	     M-PRIOQUEUE	|	   99.17% of 600	|
+|	        M-RBTREE	|	   99.04% of 937	|
+|	    M-SERIAL-BIN	|	   70.87% of 254	|
+|	   M-SERIAL-JSON	|	  100.00% of 494	|
+|	        M-SHARED	|	   99.07% of 324	|
+|	      M-SNAPSHOT	|	   99.54% of 439	|
+|	        M-STRING	|	   99.40% of 831	|
+|	          M-TREE	|	  98.46% of 1232	|
+|	         M-TUPLE	|	  100.00% of 186	|
+|	       M-VARIANT	|	   99.25% of 266	|
+|	        M-WORKER	|	    98.73% of 79	|