Added optional hash table `indexmap`

[ghost]

First of all, thank you for this dependency and all your hard work on it.

Personally I use IndexMap instead of BTreeMap for my projects, so I think it would be nice to allow people to choose IndexMap as their preferred hash table for conversion.

An example of why I use IndexMap https://github.com/bluss/indexmap/issues/269

test a_new_btreemap       ... bench:           1 ns/iter (+/- 0)
test a_new_indexmap       ... bench:           3 ns/iter (+/- 0)

test b_insert_btreemap    ... bench:  60,621,040 ns/iter (+/- 15,248,541)
test b_insert_indexmap    ... bench:  42,826,040 ns/iter (+/- 1,214,166)

test c_clone_btreemap     ... bench:  41,720,370 ns/iter (+/- 1,932,831)
test c_clone_indexmap     ... bench:  39,500,870 ns/iter (+/- 1,026,400)

test d_for_each_btreemap  ... bench:   8,487,320 ns/iter (+/- 112,188)
test d_for_each_indexmap  ... bench:   7,427,390 ns/iter (+/- 177,412)

test e_find_btreemap      ... bench:          16 ns/iter (+/- 1)
test e_find_indexmap      ... bench:          13 ns/iter (+/- 0)

test f_rfind_btreemap     ... bench:          23 ns/iter (+/- 1)
test f_rfind_indexmap     ... bench:          14 ns/iter (+/- 0)

//Edit I also updated all dependencies directly

[Diggsey]

Hi @ThisAccountHasBeenSuspended - ijson always uses its own representation internally, this is just a conversion.

With that in mind, the new feature should not disable the existing conversion from BTreeMap - when the indexmap feature flag is enabled the new conversion should exist in addition to the old conversion.

[Diggsey]

Thanks 👍