High memory usage

[guynan]

Memory usage is a bit high.

I want it to be such that only if the feature is used that the memory is allocated for said feature. I also do not want to pollute the vertex structure - thus more features being exported into the meta structure such that only if you are using more niche features do you really need to allocate that memory.

I also want to have the adjacency array to be initialised as you need it, and not for n vertices; perhaps in n/4 increments, using realloc each time to ensure that memory usage on sparse graphs is not ridiculously high.

[guynan]

I have made it such that memory is initially allocated in small blocks and if said block is exhausted, then allocates a block of similarly sized memory. Though this means that the initial outset of the memory is still high on large data sets.

I may swap this for something that starts out at a small value, and then subsequently provisions memory in multiples of the past value. This however means that there is a lower initial footprint, but memory can balloon precipitously.

[guynan]

• The vertex structure used to be 64 bytes in size to store 45 bytes
  of information, making a whopping 19 bytes (or ~30%) of
  the structure consisting of padding.

• The structure has been reordered to allow for better alignment
  properties allowing the structure to shrink down into 48 bytes
  now meaning only 3 bytes (~6%) are used for padding, this means
  memory consumption for heap allocation of vertices has shrunk
  noticably.