Degraded performance wkt.Unmarshal since v.0.9.1

[doqtri]

Hi,

I've noticed the degraded performance of the method wkt.Unmarshal since v.0.9.1
Here is the comparison

v.0.9.0

BenchmarkUnmarshall_V90-10
BenchmarkUnmarshall_V90-10    	   55581	     21352 ns/op	   11384 B/op	     101 allocs/op

v.0.9.1

BenchmarkUnmarshall_V91-10
BenchmarkUnmarshall_V91-10    	    9172	    121703 ns/op	   17722 B/op	     180 allocs/op

[paulmach]

looks like

• Replaced strings split with splitting by regexp #128

and then

• encoding/wkt: better handling/validation of missing parens #131

There were a few issues about validation that these addressed. PRs welcome.

[paulmach]

performance has been improved to much greater than v0.9.0 levels in #142 See benchmarks in that PR.

The exception is very small geometries with multiple parts (multi*, polygon). The splitting of the parts still uses the regexp. However, in most cases the improvements to the "point parsing" will greatly out-way any issues.