Optimize wildcard search executions

[jlourenc]

This PR makes wildcard searches significantly faster by reducing heap allocations where possible.

searchStrict logic has been encapsulated in a function that returns interface{} instead of *Container as the first returned parameter. This avoids the unnecessary wrapping/unwrapping.

Also, when handling a wildcard as the last path segment, array values are no longer copied. Instead the array is returned directly and then wrapped in a Container. This removes another allocation.

BenchmarkWildcardSearch has been added to assess the improvements.

Before the change:

BenchmarkWildcardSearch-16 2838511 413.2 ns/op 240 B/op 12 allocs/op

After the change:

BenchmarkWildcardSearch-16 5705732 201.8 ns/op 112 B/op 5 allocs/op