Fixed a problem where an enabled cache wouldn't start to fill up when adding data for a new index if multiple new indexes were created in the same session

Conflicts:

	kernel/src/docs/ops/cache.txt

Conflicts:

	kernel/src/docs/ops/cache.txt

Fixed potential problem with cache size updating in wrong way with concurrent updateSize together with remove or put.
(cherry picked from commit 7c7422d643630fadc6d3246a0eff0d1dfce947be)

* Updates size after shrinking of relationship id arrays, fixes a bug where the perceived size
  vs actual size diverged over time.
* The atomic array cache (cache_type=array) is now the default cache.
(cherry picked from commit f4b494b8fe871cd65f931f2018a2e17fe9816fbc)

Conflicts:

	kernel/src/main/java/org/neo4j/kernel/AbstractGraphDatabase.java

Better defaults for AtomicArrayCache

1.6 maint updates

Changed default cache type back to soft.

Merge of gc resistant cache into 1.6-maint

o Made soft reference cache default

Fixing the batch requests header issues

Conflicts:

	server/src/main/java/org/neo4j/server/rest/web/DatabaseActions.java

(cherry picked from commit 2814eee6fac2b1f5d7c17dbbc1a49273ff50a463)

Conflicts:

	kernel/src/main/java/org/neo4j/kernel/impl/transaction/ReadOnlyTxManager.java
	kernel/src/main/java/org/neo4j/kernel/impl/transaction/TxManager.java

(cherry picked from commit 84648b6d86be1e4c6895fa56123d67e67fa50503)

'dbinfo' command able to print array values too, and everything comes back as JSON for convenience if used from scripts or similar
(cherry picked from commit dbce7da4e64df793ffa6434a43b59334751bd4f4)

Increased the cache find threshold because it's still cheaper than reading loads of entries from the log
(cherry picked from commit 299e949353fbf888e9a82faa9ce19cbd5f020a52)

Added test for Start entries checksums equality and in that process fixed an issue with the current implementation
(cherry picked from commit 9623fbfa7a978b22408ec0b4152741a921052157)

Conflicts:

	kernel/src/test/java/org/neo4j/kernel/impl/transaction/xaframework/TestTxEntries.java

Uses a combination of Xid and masterId/myId as checksum instead of just timeWritten to make a transaction start record truely unique. This will allow for better branch detection
(cherry picked from commit 4489cf7be4205782c54edf98a1bd84144e1d3d13)

(cherry picked from commit 4cc35ad05a324cd45c08a776a3845c07e7bb7d7f)

(cherry picked from commit bd18c3cbda4a22196b521317959b7668422a8052)