SQLite amalgamation to 3.8.9

[denji]

The new release also conducted performance optimization of full-text queries fts3/fts4, used in combination with the OR operator. Fixed a potential integer overflow in functions sqlite3_blob_read() and sqlite3_blob_write().

Before (go test -bench=. -timeout=30s):

BenchmarkExec           200000     107113 req/s
BenchmarkQuery           50000      34443 req/s
BenchmarkParams          50000      29273 req/s
BenchmarkStmt           100000      49890 req/s
BenchmarkRows             2000       1059 req/s
BenchmarkStmtRows         2000       1088 req/s
PASS
ok      _/tmp/go/go-sqlite3-master  16.157s

After:

BenchmarkExec           200000     110548 req/s
BenchmarkQuery           50000      35105 req/s
BenchmarkParams          50000      29825 req/s
BenchmarkStmt           100000      51111 req/s
BenchmarkRows             2000       1076 req/s
BenchmarkStmtRows         2000       1100 req/s
PASS
ok      _/tmp/go/go-sqlite3-sqlite388   15.905s

[coveralls]

Coverage remained the same at 84.01% when pulling 5b6351e on denji:sqlite388 into 25d045f on mattn:master.

[coveralls]

Coverage remained the same at 84.01% when pulling 027bc7d on denji:sqlite388 into 25d045f on mattn:master.

[mattn]

Thank you. But currently, golang have issue about this.

golang/go#9356

[coveralls]

Coverage remained the same at 84.48% when pulling 48b2bc4 on denji:sqlite388 into e28cd44 on mattn:master.

[coveralls]

Coverage remained the same at 84.48% when pulling 84386b7 on denji:sqlite388 into e28cd44 on mattn:master.

[coveralls]

Coverage remained the same at 84.48% when pulling 9acd35d on denji:sqlite388 into e28cd44 on mattn:master.

[mattn]

I didn't merge this PR, but I could decide to merge latest amalgamation code to latest. thanks!
#264