Improve rendering speed by introducing dynamic MapTree cache (+Show cities in iOS)

[vshcherb]

• Today when we render the file we only use static cache using High level MapTree (top level)
• This leads that every time for tile whole map section scanned (i.e. reading header, skipping bytes, reading another header, skipping bytes...)
• Instead we could introduce MapTree cache, so once Header read it's stored in cache and next neighboor tile rendering will use that cache and operate this can in Memory
• This should increase performance dramatically on low I/O reading devices (where skip is not efficient)

Spillover

• Fix iOS show city in search show city next to search result #2993

Challenges

• We use multithreading so we should caches roughly 1 per each thread (it's ok to have copy of caches than concurrent collections)
• We need to reuse these caches (so pool is needed)
• This cache is needed for search as well, so every next search will work faster

[vshcherb]

osmandapp/OsmAnd-core#787

[vshcherb]

• Read cache prototype #21228
• Use cache during OBF section reading OsmAnd-core#799
• Test read cache for map files OsmAnd-tools#977

[vshcherb]

Testing results show no big improvement

Cache 7 depth
0.277 sec, 222 blocks total 1 858 560 bytes - 20821 objects
0.030 sec, 68 blocks total 988 160 bytes - 20821 objects
0.028 sec, 68 blocks total 988 160 bytes - 20821 objects
0.029 sec, 68 blocks total 988 160 bytes - 20821 objects

No Cache
0.278 sec, 222 blocks total 1 858 560 bytes - 20821 objects
0.033 sec, 215 blocks total 1 740 800 bytes - 20821 objects
0.031 sec, 215 blocks total 1 740 800 bytes - 20821 objects
0.033 sec, 215 blocks total 1 740 800 bytes - 20821 objects