fix identity map

[Roger-luo]

MWE: when there is a LinkedList{Any}, the following will error:

ex = :(a+b+2 * 2 + foo(2))
tl = list(ex.args...)
map(x->x, tl)

This PR will fix it, and will allow more generic (each node has different generic types) linked list.

[codecov]

Codecov Report

Merging #470 into master will decrease coverage by 31.28%.
The diff coverage is 100%.

@@             Coverage Diff             @@
##           master     #470       +/-   ##
===========================================
- Coverage    98.8%   67.52%   -31.29%     
===========================================
  Files          27       28        +1     
  Lines        1336     2069      +733     
===========================================
+ Hits         1320     1397       +77     
- Misses         16      672      +656

Impacted Files	Coverage Δ
src/list.jl	93.84% <100%> (-6.16%)	⬇️
src/container_loops.jl	35.1% <0%> (-64.9%)	⬇️
src/balanced_tree.jl	45.08% <0%> (-54.92%)	⬇️
src/delegate.jl	50% <0%> (-50%)	⬇️
src/dict_support.jl	50% <0%> (-50%)	⬇️
src/heaps/mutable_binary_heap.jl	50% <0%> (-50%)	⬇️
src/deque.jl	57.35% <0%> (-42.65%)	⬇️
src/tokens2.jl	64.1% <0%> (-35.9%)	⬇️
src/sorted_set.jl	65.24% <0%> (-34.76%)	⬇️
src/heaps/binary_heap.jl	68.25% <0%> (-31.75%)	⬇️
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 31044ef...3d9e34e. Read the comment docs.

[oxinabox]

Please give this @testset a real name.

The fact that the existing testsets do not have real names is not a convention of this package.
It is a flaw resulting from the package being older than testsets.
And noone having gotten around to renaming things with meaninful names yet

[Roger-luo]

I see. Fixed.

[oxinabox]

I actually spent a while thinking about this.
I concluded that it will do.
There is probably something smarter we can
do but this is definitely an improvement.

If you can fix the testset name we can merge

[Roger-luo]

Yeah, I was thinking to have a more elegant solution for this, e.g keep the interface, but implement several different types (for different situation), like mutable, immutable, etc.

But this is just a straight forward fix to make things work first.

[oxinabox]

Something like what @andyferris did here for determining and progressively widening the type
seems smartest.
JuliaLang/julia#30286

(I just have that PR handly for unique handy, there is something like that for Base.map(f, ::Array) I am sure.)