Testbed: 32 bit memory layout for buffer

[jerch]

Testbed PR to be able to identify differences between 32 and 16 bit memory layout for the buffer in runtime and memory usage.

Implements:

• 32 bit buffer layout as discussed in Buffer performance improvements #791
• utf8 input encoding
• direct buffer access in InputHandler.print (therefore the JS array buffer will not work here)

Buffer layout:

|             uint32_t             |        uint32_t         |        uint32_t         |
|             `content`            |          `FG`           |          `BG`           |
| wcwidth(2) comb(1) codepoint(21) | flags(8) R(8) G(8) B(8) | flags(8) R(8) G(8) B(8) |

Else its based on current master, additionally it includes the webgl renderer as default (makes it easier to spot buffer related things in the timeline).
Simply run the demo app and compare the timeline results with the 16 bit variant #1855.

Note: To see the impact of the utf8 input handling alone, the timeline results can be compared to the webgl PR branch.

/cc @Tyriar, @mofux, @bgw

Edit: The tests will not pass, I did not apply the change all over the codebase, just to the critical parts. If the build does not finish, run npm run watch and npm run watch-addons instead.

[jerch]

First results (tested with ls -lR /usr/lib in demo app with default settings):

[jerch]

Results in xterm-benchmark:

            Case "#1" : 10 runs - average runtime: 1545.61 ms
            Case "#1" : 10 runs - average throughput: 31.27 MB/s

[Tyriar]

I think this is definitely the way to go to start with, double the (small amount of) memory for a little faster runtime and it's way more simple especially since we don't have to deal with how attributes are stored and invalidated.

[Tyriar]

@jerch let's move the memory layout discussion over to #484, I think we're in agreement for how the buffer should work initially and it's sounding like we can probably get true color going in 3.11 if we go with this lower risk approach?

[jerch]

@Tyriar Yes, with this approach here the extended attrs are much easier and less error prone to implement, prolly can be done in one true color PR.

[jerch]

Closing the PR as I think we all agree to go with this approach for now.