Restore CUDA feature flag in build.rs

[winston-h-zhang]

No description provided.

[huitseeker]

Measured effect on Lurk at 2efe6ec44:

LEM Fibonacci Prove - rc = 100/fib/num-100-2efe6ec44becd5ec4f572ec5ed89f9f10b454a82-2024-01-08
                        time:   [1.3899 s 1.3941 s 1.3985 s]
                        thrpt:  [572.03  elem/s 573.84  elem/s 575.57  elem/s]
                 change:
                        time:   [-71.368% -71.229% -71.100%] (p = 0.00 < 0.05)
                        thrpt:  [+246.02% +247.58% +249.26%]
                        Performance has improved.
LEM Fibonacci Prove - rc = 100/fib/num-200-2efe6ec44becd5ec4f572ec5ed89f9f10b454a82-2024-01-08
                        time:   [2.6659 s 2.6844 s 2.7079 s]
                        thrpt:  [553.93  elem/s 558.78  elem/s 562.65  elem/s]
                 change:
                        time:   [-72.715% -72.437% -72.154%] (p = 0.00 < 0.05)
                        thrpt:  [+259.12% +262.81% +266.50%]
                        Performance has improved.

LEM Fibonacci Prove - rc = 600/fib/num-100-2efe6ec44becd5ec4f572ec5ed89f9f10b454a82-2024-01-08
                        time:   [1.5681 s 1.5699 s 1.5718 s]
                        thrpt:  [763.47  elem/s 764.39  elem/s 765.26  elem/s]
                 change:
                        time:   [-63.885% -62.815% -61.891%] (p = 0.00 < 0.05)
                        thrpt:  [+162.40% +168.93% +176.89%]
                        Performance has improved.
LEM Fibonacci Prove - rc = 600/fib/num-200-2efe6ec44becd5ec4f572ec5ed89f9f10b454a82-2024-01-08
                        time:   [2.6534 s 2.6614 s 2.6723 s]
                        thrpt:  [673.59  elem/s 676.34  elem/s 678.36  elem/s]
                 change:
                        time:   [-62.563% -62.117% -61.598%] (p = 0.00 < 0.05)
                        thrpt:  [+160.41% +163.97% +167.12%]
                        Performance has improved.