This adds at least one memory allocation (`collect` into a `Vec`) but is still faster for a reasonable number of IDs.

There is a performance penalty (+13%) but this collection enables cheap
deletions.

When creating an index (`IndexedReader::create_index`) the unused `Blob`s are not copied into memory, which really improves performance for this operation (5x faster).

This allows filtering elements and finding their dependencies by
decoding as few blobs as possible.

for new protobuf version 2.10.1

The version of compiled files needs to match the protobuf version,
including patch level.

for new protobuf version 2.10.0

to avoid confusion. The returned slice needs delta decoding to retrieve
the node ids.

because of lazy_static.

and fix some clippy warnings.

because the builds are failing with protobuf 1.7