Sparse Bayes (#1748)

* Sparse Bayes

* updated perf test

* improved tests with real data

* Making sparse works in incremental mapping

* bookkeeping optimization

* small opt

* refactoring

* splitting dense and sparse in different classes to make the code more lisible

* cleanup comments

* fixing CI

* Making all Bayes tests testing both dense and sparse

* Added multisession_3it integration test (test memory management, multisession and dense/sparse bayes in that settings)

* optimized sparse when transfer/retrieval happens (was slower than dense for that case)

* Testing retrieval param variants

* Updated multisession_3it integration tests to compare loop closure hypotheses

* bump version

* Fixed ui sum of prediction

* adding g2o gtsam to linux ci

* cleanup

* added debug crash log for ci

* Simplified Bayes/SparsePrediction description

* Dont show too dense for sparse on small maps (e.g., when we just started a new map)

* fixing amd64v3 issue with gtsam on ci ubuntu 26

* Dot not auto switch to dense based on map size.

* updating test range

* added coverage tests

* Adressing coverage

* ignore one line in coverage for purpose

corelib/test/CMakeLists.txt

@@ -119,6 +119,24 @@ IF(BUILD_PERF_TESTS)
 	set_tests_properties(test_flann_index_perf PROPERTIES
 			TIMEOUT ${_perf_timeout}
 			LABELS  "performance")
+
+	# Comparison of the dense and the sparse prediction x posterior multiplication of
+	# BayesFilter (Bayes/SparsePrediction), against the size of the map, how connected
+	# its graph is and the depth of the prediction model. Over synthetic graphs, and over
+	# the graph of a real map read from data/tests/large_reduced_graph.g2o, whose link
+	# types decide how much of the map a column of the prediction holds:
+	#   bin/test_bayesfilter_perf
+	#   bin/test_bayesfilter_perf --gtest_filter=-*LargeMap*:-*RealMap*
+	# Its own executable: it spends its time benchmarking rather than asserting, and the
+	# largest maps it builds allocate a gigabyte for the dense prediction matrix,
+	# which in a unit test shard would look like a leak.
+	add_executable(test_bayesfilter_perf perf_bayesfilter.cpp)
+	target_link_libraries(test_bayesfilter_perf gtest_main rtabmap_core)
+
+	add_test(NAME test_bayesfilter_perf COMMAND test_bayesfilter_perf)
+	set_tests_properties(test_bayesfilter_perf PROPERTIES
+			TIMEOUT ${_perf_timeout}
+			LABELS  "performance")
 ENDIF(BUILD_PERF_TESTS)
 
 # Rtabmap end-to-end replay of sample DBs (test data fetched by