Implement Turing-like model macro

[sunxd3]

Address #281

Edited:
I wrote some documentation in markdown for this PR and design, there is a slight change from last weeks' Monday meeting.

The markdown doc should be concise enough and I don't think I can make it much shorter without losing essential information.

So could the reviews, please, give https://github.com/TuringLang/JuliaBUGS.jl/blob/sunxd/new_model_syntax/docs/src/julia_syntax.md a quick read?

[coveralls]

Pull Request Test Coverage Report for Build 14760776262

Details

• 74 of 87 (85.06%) changed or added relevant lines in 2 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+0.1%) to 83.512%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/JuliaBUGS.jl	10	11	90.91%
src/model_macro.jl	64	76	84.21%

Files with Coverage Reduction	New Missed Lines	%
src/JuliaBUGS.jl	1	83.72%

Totals
Change from base Build 14757438446:	0.1%
Covered Lines:	2107
Relevant Lines:	2523

---

💛 - Coveralls

[mhauru]

I read the Markdown, it's well written. There's a broader conversation about @parameters that concerns DPPL too, as discussed e.g. today, but I assume that's outside the scope of this PR. I didn't actually read the implementation of the macros.

I added a couple of suggested comments to explain the tests, do edit them if I they are wrong.

[github-actions]

Benchmark results on macOS (aarch64)

BridgeStan not found at location specified by $BRIDGESTAN environment variable, downloading version 2.6.2 to /Users/runner/.bridgestan/bridgestan-2.6.2
Done!

Stan results:

Model	Parameters	Density Time (µs)	Density+Gradient Time (µs)
rats	65	5.0166	7.06933
pumps	12	0.8262	0.958345
bones	13	74.0	84.375
oxford	244	10.7705	12.75
epil	303	31.125	35.75
lsat	1006	97.541	133.125
schools	133	224.313	341.25
beetles	2	0.740625	0.803912
air	5	0.817139	0.902269

JuliaBUGS Mooncake results:

Model	Parameters	Density Time (µs)	Density+Gradient Time (µs)
rats	65	1.25191	7.81933
pumps	12	0.512207	1.95836
bones	33	60.375	264.625
oxford	244	21.917	55.584
epil	303	7.472	37.25
lsat	1006	54.875	243.23
schools	133	86.292	704.875
beetles	2	1.15332	4.175
air	5	0.310484	1.28239

---

Benchmark results on Ubuntu (x64)

BridgeStan not found at location specified by $BRIDGESTAN environment variable, downloading version 2.6.2 to /home/runner/.bridgestan/bridgestan-2.6.2
Done!

Stan results:

Model	Parameters	Density Time (µs)	Density+Gradient Time (µs)
rats	65	5.2158	7.60733
pumps	12	0.952103	1.21487
bones	13	69.95	86.29
oxford	244	13.721	17.733
epil	303	28.934	35.385
lsat	1006	164.315	199.421
schools	133	485.632	720.498
beetles	2	0.850412	1.00428
air	5	0.667591	0.809833

JuliaBUGS Mooncake results:

Model	Parameters	Density Time (µs)	Density+Gradient Time (µs)
rats	65	1.9356	13.8005
pumps	12	0.5939	2.71873
bones	33	47.889	353.426
oxford	244	23.504	72.915
epil	303	10.79	55.072
lsat	1006	89.736	363.58
schools	133	102.766	1085.04
beetles	2	1.09742	4.57183
air	5	0.341453	1.36633

[sunxd3]

@yebai I am merging this. It is not a breaking change because the model macro is not exported and compile's behavior is not changed yet. I want this to be in the main branch to update the interop examples.

We can adjust the public interface later, and bump the version then.