RFR(s): various stringStream issues

[Thomas Stüfe]

Hi all.

this is a smallish change which rewrites stringStream to achieve two goals:

- have a buffer size cap in dynamic mode: faulty - or just very chatty -
printing code can use up lots of memory, and we have no inbuilt limit. We
should have one. This is similar to
https://bugs.openjdk.java.net/browse/JDK-8220394

- Since 8224193, stringStream uses no resource area anymore as backing
buffer but C-heap. To reduce the number of malloc calls for small instances
of this stream I added a small internal buffer to be used for small output.

I rewrote stringStream to be more in line with how the rest of our code
looks today (constness, initializer lists etc).

I also tweaked the tests a lot and believe this code pretty well tested.

JBS: https://bugs.openjdk.java.net/browse/JDK-8224213
webrev:
http://cr.openjdk.java.net/~stuefe/webrevs/8224213-stringstream-maxcap/webrev.00/webrev/index.html

Thank you, Thomas